Chapter 5: Sampling DistributionsChapter 5: Sampling Distributions Tuesday (3/10/15): 5.1 {Sampling...
Transcript of Chapter 5: Sampling DistributionsChapter 5: Sampling Distributions Tuesday (3/10/15): 5.1 {Sampling...
![Page 1: Chapter 5: Sampling DistributionsChapter 5: Sampling Distributions Tuesday (3/10/15): 5.1 {Sampling Distributions {Sample means and the Law of Large Numbers {The Central Limit Theorem](https://reader030.fdocuments.us/reader030/viewer/2022040223/5e5413034285d826427cbc6a/html5/thumbnails/1.jpg)
Chapter 5: Sampling Distributions
Tuesday (3/10/15): 5.1
– Sampling Distributions
– Sample means and the Law of Large Numbers
– The Central Limit Theorem
Thursday (3/12/15): 5.2, Sampling Distributions for Counts andProportionsHomework: WebAssign due Friday, Excel assignment posted today, dueTuesday after Spring Break.
March 12, 2015 1 / 28
![Page 2: Chapter 5: Sampling DistributionsChapter 5: Sampling Distributions Tuesday (3/10/15): 5.1 {Sampling Distributions {Sample means and the Law of Large Numbers {The Central Limit Theorem](https://reader030.fdocuments.us/reader030/viewer/2022040223/5e5413034285d826427cbc6a/html5/thumbnails/2.jpg)
Parameters and Statistics
Recall:
– A parameter is a number that describes an actual characteristic of apopulation. Assuming we cannot sample the entire population,parameters are never known exactly.
– A statistic is a number that describes a characteristic of a sample.
We use statistics to estimate parameters.
Notation for samples vs. populations:
Symbol Use
µ Population mean
x̄ Sample mean
σ Population standard deviation
s, sx Sample standard deviation
March 12, 2015 2 / 28
![Page 3: Chapter 5: Sampling DistributionsChapter 5: Sampling Distributions Tuesday (3/10/15): 5.1 {Sampling Distributions {Sample means and the Law of Large Numbers {The Central Limit Theorem](https://reader030.fdocuments.us/reader030/viewer/2022040223/5e5413034285d826427cbc6a/html5/thumbnails/3.jpg)
Parameters and Statistics
Recall:
– A parameter is a number that describes an actual characteristic of apopulation. Assuming we cannot sample the entire population,parameters are never known exactly.
– A statistic is a number that describes a characteristic of a sample.
We use statistics to estimate parameters.
Notation for samples vs. populations:
Symbol Use
µ Population mean
x̄ Sample mean
σ Population standard deviation
s, sx Sample standard deviation
March 12, 2015 2 / 28
![Page 4: Chapter 5: Sampling DistributionsChapter 5: Sampling Distributions Tuesday (3/10/15): 5.1 {Sampling Distributions {Sample means and the Law of Large Numbers {The Central Limit Theorem](https://reader030.fdocuments.us/reader030/viewer/2022040223/5e5413034285d826427cbc6a/html5/thumbnails/4.jpg)
Parameters and Statistics
Recall:
– A parameter is a number that describes an actual characteristic of apopulation. Assuming we cannot sample the entire population,parameters are never known exactly.
– A statistic is a number that describes a characteristic of a sample.
We use statistics to estimate parameters.
Notation for samples vs. populations:
Symbol Use
µ Population mean
x̄ Sample mean
σ Population standard deviation
s, sx Sample standard deviation
March 12, 2015 2 / 28
![Page 5: Chapter 5: Sampling DistributionsChapter 5: Sampling Distributions Tuesday (3/10/15): 5.1 {Sampling Distributions {Sample means and the Law of Large Numbers {The Central Limit Theorem](https://reader030.fdocuments.us/reader030/viewer/2022040223/5e5413034285d826427cbc6a/html5/thumbnails/5.jpg)
Why sampling distributions?
Fundamental idea:
We want to use sample statistics to estimate population parameters.Different random samples will yield different statistics, so we must
know the distribution of these sample statistics!
– Sample statistics will be treated like random variables, and wealready know how to find the distribution (PDF) of a random variable(chapter 4).– Recall: given a random variable X, its probability distribution is atable (for discrete RV’s) or function (for continuous RV’s) thatprovides the range of possible outputs of X and their probabilities.
X 2 4 8
pX(x) 1/4 1/2 1/4
March 12, 2015 3 / 28
![Page 6: Chapter 5: Sampling DistributionsChapter 5: Sampling Distributions Tuesday (3/10/15): 5.1 {Sampling Distributions {Sample means and the Law of Large Numbers {The Central Limit Theorem](https://reader030.fdocuments.us/reader030/viewer/2022040223/5e5413034285d826427cbc6a/html5/thumbnails/6.jpg)
Why sampling distributions?
Fundamental idea:
We want to use sample statistics to estimate population parameters.Different random samples will yield different statistics, so we must
know the distribution of these sample statistics!
– Sample statistics will be treated like random variables, and wealready know how to find the distribution (PDF) of a random variable(chapter 4).
– Recall: given a random variable X, its probability distribution is atable (for discrete RV’s) or function (for continuous RV’s) thatprovides the range of possible outputs of X and their probabilities.
X 2 4 8
pX(x) 1/4 1/2 1/4
March 12, 2015 3 / 28
![Page 7: Chapter 5: Sampling DistributionsChapter 5: Sampling Distributions Tuesday (3/10/15): 5.1 {Sampling Distributions {Sample means and the Law of Large Numbers {The Central Limit Theorem](https://reader030.fdocuments.us/reader030/viewer/2022040223/5e5413034285d826427cbc6a/html5/thumbnails/7.jpg)
Why sampling distributions?
Fundamental idea:
We want to use sample statistics to estimate population parameters.Different random samples will yield different statistics, so we must
know the distribution of these sample statistics!
– Sample statistics will be treated like random variables, and wealready know how to find the distribution (PDF) of a random variable(chapter 4).– Recall: given a random variable X, its probability distribution is atable (for discrete RV’s) or function (for continuous RV’s) thatprovides the range of possible outputs of X and their probabilities.
X 2 4 8
pX(x) 1/4 1/2 1/4
March 12, 2015 3 / 28
![Page 8: Chapter 5: Sampling DistributionsChapter 5: Sampling Distributions Tuesday (3/10/15): 5.1 {Sampling Distributions {Sample means and the Law of Large Numbers {The Central Limit Theorem](https://reader030.fdocuments.us/reader030/viewer/2022040223/5e5413034285d826427cbc6a/html5/thumbnails/8.jpg)
Sampling Example
Example: suppose there are 50,000 students at UA. We give everyone aslip of paper with a number written on it. 10,000 students get a 1,10,000 students get a 2, etc, up to 5.
– If we select a student at random and look at their number, this is arandom variable X.– What is the distribution of X?
X 1 2 3 4 5
pX(x) 1/5 1/5 1/5 1/5 1/5
– What is the population mean of X?
µX =
5∑i=1
pixi =1
5(1 + 2 + 3 + 4 + 5) = 2.5
– What is the population standard deviation?
σX =
5∑i=1
pi(xi − µX)2 =1
5
((1− 2.5)2 + . . .+ (5− 2.5)2
)= 2.25
March 12, 2015 4 / 28
![Page 9: Chapter 5: Sampling DistributionsChapter 5: Sampling Distributions Tuesday (3/10/15): 5.1 {Sampling Distributions {Sample means and the Law of Large Numbers {The Central Limit Theorem](https://reader030.fdocuments.us/reader030/viewer/2022040223/5e5413034285d826427cbc6a/html5/thumbnails/9.jpg)
Sampling Example
Example: suppose there are 50,000 students at UA. We give everyone aslip of paper with a number written on it. 10,000 students get a 1,10,000 students get a 2, etc, up to 5.– If we select a student at random and look at their number, this is arandom variable X.
– What is the distribution of X?
X 1 2 3 4 5
pX(x) 1/5 1/5 1/5 1/5 1/5
– What is the population mean of X?
µX =
5∑i=1
pixi =1
5(1 + 2 + 3 + 4 + 5) = 2.5
– What is the population standard deviation?
σX =
5∑i=1
pi(xi − µX)2 =1
5
((1− 2.5)2 + . . .+ (5− 2.5)2
)= 2.25
March 12, 2015 4 / 28
![Page 10: Chapter 5: Sampling DistributionsChapter 5: Sampling Distributions Tuesday (3/10/15): 5.1 {Sampling Distributions {Sample means and the Law of Large Numbers {The Central Limit Theorem](https://reader030.fdocuments.us/reader030/viewer/2022040223/5e5413034285d826427cbc6a/html5/thumbnails/10.jpg)
Sampling Example
Example: suppose there are 50,000 students at UA. We give everyone aslip of paper with a number written on it. 10,000 students get a 1,10,000 students get a 2, etc, up to 5.– If we select a student at random and look at their number, this is arandom variable X.– What is the distribution of X?
X 1 2 3 4 5
pX(x) 1/5 1/5 1/5 1/5 1/5
– What is the population mean of X?
µX =
5∑i=1
pixi =1
5(1 + 2 + 3 + 4 + 5) = 2.5
– What is the population standard deviation?
σX =
5∑i=1
pi(xi − µX)2 =1
5
((1− 2.5)2 + . . .+ (5− 2.5)2
)= 2.25
March 12, 2015 4 / 28
![Page 11: Chapter 5: Sampling DistributionsChapter 5: Sampling Distributions Tuesday (3/10/15): 5.1 {Sampling Distributions {Sample means and the Law of Large Numbers {The Central Limit Theorem](https://reader030.fdocuments.us/reader030/viewer/2022040223/5e5413034285d826427cbc6a/html5/thumbnails/11.jpg)
Sampling Example
Example: suppose there are 50,000 students at UA. We give everyone aslip of paper with a number written on it. 10,000 students get a 1,10,000 students get a 2, etc, up to 5.– If we select a student at random and look at their number, this is arandom variable X.– What is the distribution of X?
X 1 2 3 4 5
pX(x) 1/5 1/5 1/5 1/5 1/5
– What is the population mean of X?
µX =
5∑i=1
pixi =1
5(1 + 2 + 3 + 4 + 5) = 2.5
– What is the population standard deviation?
σX =
5∑i=1
pi(xi − µX)2 =1
5
((1− 2.5)2 + . . .+ (5− 2.5)2
)= 2.25
March 12, 2015 4 / 28
![Page 12: Chapter 5: Sampling DistributionsChapter 5: Sampling Distributions Tuesday (3/10/15): 5.1 {Sampling Distributions {Sample means and the Law of Large Numbers {The Central Limit Theorem](https://reader030.fdocuments.us/reader030/viewer/2022040223/5e5413034285d826427cbc6a/html5/thumbnails/12.jpg)
Sampling Example
Example: suppose there are 50,000 students at UA. We give everyone aslip of paper with a number written on it. 10,000 students get a 1,10,000 students get a 2, etc, up to 5.– If we select a student at random and look at their number, this is arandom variable X.– What is the distribution of X?
X 1 2 3 4 5
pX(x) 1/5 1/5 1/5 1/5 1/5
– What is the population mean of X?
µX =
5∑i=1
pixi =1
5(1 + 2 + 3 + 4 + 5) = 2.5
– What is the population standard deviation?
σX =
5∑i=1
pi(xi − µX)2 =1
5
((1− 2.5)2 + . . .+ (5− 2.5)2
)= 2.25
March 12, 2015 4 / 28
![Page 13: Chapter 5: Sampling DistributionsChapter 5: Sampling Distributions Tuesday (3/10/15): 5.1 {Sampling Distributions {Sample means and the Law of Large Numbers {The Central Limit Theorem](https://reader030.fdocuments.us/reader030/viewer/2022040223/5e5413034285d826427cbc6a/html5/thumbnails/13.jpg)
Sample Means and the Law of Large Numbers
– Now, suppose we select 100 students at random and record theirnumber. You can simulate this in excel using RANDBETWEEN(1,5).
1, 5, 2, 2, 5, 4, 3, 1, 3, 5, . . .
– We think of each of the above numbers as a random variable Xi. Inother words, we have 100 random variables!– We can compute the sample mean x̄ = 1
n
∑ni=1Xi, with n = 100.
Law of large numbers: as the number of samples (n) gets large,x̄→ µX .
– We are more interested in the variability of x̄ across allpossible samples of size 100.– In particular, x̄ is a random variable, since it is a linear combinationof random variables! What is its distribution?
March 12, 2015 5 / 28
![Page 14: Chapter 5: Sampling DistributionsChapter 5: Sampling Distributions Tuesday (3/10/15): 5.1 {Sampling Distributions {Sample means and the Law of Large Numbers {The Central Limit Theorem](https://reader030.fdocuments.us/reader030/viewer/2022040223/5e5413034285d826427cbc6a/html5/thumbnails/14.jpg)
Sample Means and the Law of Large Numbers
– Now, suppose we select 100 students at random and record theirnumber. You can simulate this in excel using RANDBETWEEN(1,5).
1, 5, 2, 2, 5, 4, 3, 1, 3, 5, . . .
– We think of each of the above numbers as a random variable Xi. Inother words, we have 100 random variables!
– We can compute the sample mean x̄ = 1n
∑ni=1Xi, with n = 100.
Law of large numbers: as the number of samples (n) gets large,x̄→ µX .
– We are more interested in the variability of x̄ across allpossible samples of size 100.– In particular, x̄ is a random variable, since it is a linear combinationof random variables! What is its distribution?
March 12, 2015 5 / 28
![Page 15: Chapter 5: Sampling DistributionsChapter 5: Sampling Distributions Tuesday (3/10/15): 5.1 {Sampling Distributions {Sample means and the Law of Large Numbers {The Central Limit Theorem](https://reader030.fdocuments.us/reader030/viewer/2022040223/5e5413034285d826427cbc6a/html5/thumbnails/15.jpg)
Sample Means and the Law of Large Numbers
– Now, suppose we select 100 students at random and record theirnumber. You can simulate this in excel using RANDBETWEEN(1,5).
1, 5, 2, 2, 5, 4, 3, 1, 3, 5, . . .
– We think of each of the above numbers as a random variable Xi. Inother words, we have 100 random variables!– We can compute the sample mean x̄ = 1
n
∑ni=1Xi, with n = 100.
Law of large numbers: as the number of samples (n) gets large,x̄→ µX .
– We are more interested in the variability of x̄ across allpossible samples of size 100.– In particular, x̄ is a random variable, since it is a linear combinationof random variables! What is its distribution?
March 12, 2015 5 / 28
![Page 16: Chapter 5: Sampling DistributionsChapter 5: Sampling Distributions Tuesday (3/10/15): 5.1 {Sampling Distributions {Sample means and the Law of Large Numbers {The Central Limit Theorem](https://reader030.fdocuments.us/reader030/viewer/2022040223/5e5413034285d826427cbc6a/html5/thumbnails/16.jpg)
Sample Means and the Law of Large Numbers
– Now, suppose we select 100 students at random and record theirnumber. You can simulate this in excel using RANDBETWEEN(1,5).
1, 5, 2, 2, 5, 4, 3, 1, 3, 5, . . .
– We think of each of the above numbers as a random variable Xi. Inother words, we have 100 random variables!– We can compute the sample mean x̄ = 1
n
∑ni=1Xi, with n = 100.
Law of large numbers: as the number of samples (n) gets large,x̄→ µX .
– We are more interested in the variability of x̄ across allpossible samples of size 100.– In particular, x̄ is a random variable, since it is a linear combinationof random variables! What is its distribution?
March 12, 2015 5 / 28
![Page 17: Chapter 5: Sampling DistributionsChapter 5: Sampling Distributions Tuesday (3/10/15): 5.1 {Sampling Distributions {Sample means and the Law of Large Numbers {The Central Limit Theorem](https://reader030.fdocuments.us/reader030/viewer/2022040223/5e5413034285d826427cbc6a/html5/thumbnails/17.jpg)
Sample Means and the Law of Large Numbers
– Now, suppose we select 100 students at random and record theirnumber. You can simulate this in excel using RANDBETWEEN(1,5).
1, 5, 2, 2, 5, 4, 3, 1, 3, 5, . . .
– We think of each of the above numbers as a random variable Xi. Inother words, we have 100 random variables!– We can compute the sample mean x̄ = 1
n
∑ni=1Xi, with n = 100.
Law of large numbers: as the number of samples (n) gets large,x̄→ µX .
– We are more interested in the variability of x̄ across allpossible samples of size 100.
– In particular, x̄ is a random variable, since it is a linear combinationof random variables! What is its distribution?
March 12, 2015 5 / 28
![Page 18: Chapter 5: Sampling DistributionsChapter 5: Sampling Distributions Tuesday (3/10/15): 5.1 {Sampling Distributions {Sample means and the Law of Large Numbers {The Central Limit Theorem](https://reader030.fdocuments.us/reader030/viewer/2022040223/5e5413034285d826427cbc6a/html5/thumbnails/18.jpg)
Sample Means and the Law of Large Numbers
– Now, suppose we select 100 students at random and record theirnumber. You can simulate this in excel using RANDBETWEEN(1,5).
1, 5, 2, 2, 5, 4, 3, 1, 3, 5, . . .
– We think of each of the above numbers as a random variable Xi. Inother words, we have 100 random variables!– We can compute the sample mean x̄ = 1
n
∑ni=1Xi, with n = 100.
Law of large numbers: as the number of samples (n) gets large,x̄→ µX .
– We are more interested in the variability of x̄ across allpossible samples of size 100.– In particular, x̄ is a random variable, since it is a linear combinationof random variables! What is its distribution?
March 12, 2015 5 / 28
![Page 19: Chapter 5: Sampling DistributionsChapter 5: Sampling Distributions Tuesday (3/10/15): 5.1 {Sampling Distributions {Sample means and the Law of Large Numbers {The Central Limit Theorem](https://reader030.fdocuments.us/reader030/viewer/2022040223/5e5413034285d826427cbc6a/html5/thumbnails/19.jpg)
Bias, Variability, Unbiased Estimators
Recall a concept from chapter 3:
If we compute a statistic (such as x̄ or sX) over all possible samples ofa certain size, this statistic will vary.
– Bias is when the mean of the statistic, over all samples, is differentfrom the population parameter. A statistic is an unbiasedestimator if its mean over all samples is actually equal to thepopulation parameter.
– Variability is how spread out the statistic is over all samples
March 12, 2015 6 / 28
![Page 20: Chapter 5: Sampling DistributionsChapter 5: Sampling Distributions Tuesday (3/10/15): 5.1 {Sampling Distributions {Sample means and the Law of Large Numbers {The Central Limit Theorem](https://reader030.fdocuments.us/reader030/viewer/2022040223/5e5413034285d826427cbc6a/html5/thumbnails/20.jpg)
Bias, Variability, Unbiased Estimators
Recall a concept from chapter 3:
If we compute a statistic (such as x̄ or sX) over all possible samples ofa certain size, this statistic will vary.
– Bias is when the mean of the statistic, over all samples, is differentfrom the population parameter. A statistic is an unbiasedestimator if its mean over all samples is actually equal to thepopulation parameter.
– Variability is how spread out the statistic is over all samples
March 12, 2015 6 / 28
![Page 21: Chapter 5: Sampling DistributionsChapter 5: Sampling Distributions Tuesday (3/10/15): 5.1 {Sampling Distributions {Sample means and the Law of Large Numbers {The Central Limit Theorem](https://reader030.fdocuments.us/reader030/viewer/2022040223/5e5413034285d826427cbc6a/html5/thumbnails/21.jpg)
Bias, Variability, Unbiased Estimators
Recall a concept from chapter 3:
If we compute a statistic (such as x̄ or sX) over all possible samples ofa certain size, this statistic will vary.
– Bias is when the mean of the statistic, over all samples, is differentfrom the population parameter. A statistic is an unbiasedestimator if its mean over all samples is actually equal to thepopulation parameter.
– Variability is how spread out the statistic is over all samples
March 12, 2015 6 / 28
![Page 22: Chapter 5: Sampling DistributionsChapter 5: Sampling Distributions Tuesday (3/10/15): 5.1 {Sampling Distributions {Sample means and the Law of Large Numbers {The Central Limit Theorem](https://reader030.fdocuments.us/reader030/viewer/2022040223/5e5413034285d826427cbc6a/html5/thumbnails/22.jpg)
Bias, Variability, Unbiased Estimators
Recall a concept from chapter 3:
If we compute a statistic (such as x̄ or sX) over all possible samples ofa certain size, this statistic will vary.
– Bias is when the mean of the statistic, over all samples, is differentfrom the population parameter. A statistic is an unbiasedestimator if its mean over all samples is actually equal to thepopulation parameter.
– Variability is how spread out the statistic is over all samples
March 12, 2015 6 / 28
![Page 23: Chapter 5: Sampling DistributionsChapter 5: Sampling Distributions Tuesday (3/10/15): 5.1 {Sampling Distributions {Sample means and the Law of Large Numbers {The Central Limit Theorem](https://reader030.fdocuments.us/reader030/viewer/2022040223/5e5413034285d826427cbc6a/html5/thumbnails/23.jpg)
Bias, Variability, Unbiased Estimators
Recall a concept from chapter 3:
If we compute a statistic (such as x̄ or sX) over all possible samples ofa certain size, this statistic will vary.
– Bias is when the mean of the statistic, over all samples, is differentfrom the population parameter. A statistic is an unbiasedestimator if its mean over all samples is actually equal to thepopulation parameter.
– Variability is how spread out the statistic is over all samples
March 12, 2015 6 / 28
![Page 24: Chapter 5: Sampling DistributionsChapter 5: Sampling Distributions Tuesday (3/10/15): 5.1 {Sampling Distributions {Sample means and the Law of Large Numbers {The Central Limit Theorem](https://reader030.fdocuments.us/reader030/viewer/2022040223/5e5413034285d826427cbc6a/html5/thumbnails/24.jpg)
Mean and standard deviation of the sample mean
Yes, that’s right - we are going to compute the mean of the mean andthe standard deviation of the mean! Here is what we mean:
– For a given sample size n, where n is small relative to thepopulation size, each individual sample is a random variable Xi
which is independent from the others and identicallydistributed. In other words, the distribution of X10 is the same asthe distribution of X1000.
– In particular, the population mean and standard deviation of each Xi
is the same:
µXi = µX = µ, σXi = σX = σ
Another way of saying this: choosing a random sample of 100without replacement from a very large population (e.g. 50,000) willresult in the same behavior as if we sampled with replacement.
March 12, 2015 7 / 28
![Page 25: Chapter 5: Sampling DistributionsChapter 5: Sampling Distributions Tuesday (3/10/15): 5.1 {Sampling Distributions {Sample means and the Law of Large Numbers {The Central Limit Theorem](https://reader030.fdocuments.us/reader030/viewer/2022040223/5e5413034285d826427cbc6a/html5/thumbnails/25.jpg)
Mean and standard deviation of the sample mean
Yes, that’s right - we are going to compute the mean of the mean andthe standard deviation of the mean! Here is what we mean:
– For a given sample size n, where n is small relative to thepopulation size, each individual sample is a random variable Xi
which is independent from the others and identicallydistributed. In other words, the distribution of X10 is the same asthe distribution of X1000.
– In particular, the population mean and standard deviation of each Xi
is the same:
µXi = µX = µ, σXi = σX = σ
Another way of saying this: choosing a random sample of 100without replacement from a very large population (e.g. 50,000) willresult in the same behavior as if we sampled with replacement.
March 12, 2015 7 / 28
![Page 26: Chapter 5: Sampling DistributionsChapter 5: Sampling Distributions Tuesday (3/10/15): 5.1 {Sampling Distributions {Sample means and the Law of Large Numbers {The Central Limit Theorem](https://reader030.fdocuments.us/reader030/viewer/2022040223/5e5413034285d826427cbc6a/html5/thumbnails/26.jpg)
Mean and standard deviation of the sample mean
Yes, that’s right - we are going to compute the mean of the mean andthe standard deviation of the mean! Here is what we mean:
– For a given sample size n, where n is small relative to thepopulation size, each individual sample is a random variable Xi
which is independent from the others and identicallydistributed. In other words, the distribution of X10 is the same asthe distribution of X1000.
– In particular, the population mean and standard deviation of each Xi
is the same:
µXi = µX = µ, σXi = σX = σ
Another way of saying this: choosing a random sample of 100without replacement from a very large population (e.g. 50,000) willresult in the same behavior as if we sampled with replacement.
March 12, 2015 7 / 28
![Page 27: Chapter 5: Sampling DistributionsChapter 5: Sampling Distributions Tuesday (3/10/15): 5.1 {Sampling Distributions {Sample means and the Law of Large Numbers {The Central Limit Theorem](https://reader030.fdocuments.us/reader030/viewer/2022040223/5e5413034285d826427cbc6a/html5/thumbnails/27.jpg)
Mean and standard deviation of the sample mean
– Knowing that µXi = µ for all i, we can compute the mean of x̄ (recallthat x̄ is a random variable!)
µx̄ = E[x̄] = E
[1
n
n∑i=1
Xi
]=
1
n
[n∑
i=1
E[Xi]
]=
1
n
n∑i=1
µ =1
nn · µ = µ
So, this says that x̄ is an unbiased estimator of µ, since the mean ofthe sampling distribution is the same as the mean of the population!
– Note: the assumption that n is small compared to the population iscrucial. If n is not small compared to population, Law of LargeNumbers is more useful.– We can also compute the sample variance/standard deviation:
σ2x̄ =
1
n2
n∑i=1
σ2Xi
=1
n2
n∑i=1
σ2 =σ2
n
so, σx̄ = σ/√n
March 12, 2015 8 / 28
![Page 28: Chapter 5: Sampling DistributionsChapter 5: Sampling Distributions Tuesday (3/10/15): 5.1 {Sampling Distributions {Sample means and the Law of Large Numbers {The Central Limit Theorem](https://reader030.fdocuments.us/reader030/viewer/2022040223/5e5413034285d826427cbc6a/html5/thumbnails/28.jpg)
Mean and standard deviation of the sample mean
– Knowing that µXi = µ for all i, we can compute the mean of x̄ (recallthat x̄ is a random variable!)
µx̄ = E[x̄] = E
[1
n
n∑i=1
Xi
]=
1
n
[n∑
i=1
E[Xi]
]=
1
n
n∑i=1
µ =1
nn · µ = µ
So, this says that x̄ is an unbiased estimator of µ, since the mean ofthe sampling distribution is the same as the mean of the population!– Note: the assumption that n is small compared to the population iscrucial. If n is not small compared to population, Law of LargeNumbers is more useful.
– We can also compute the sample variance/standard deviation:
σ2x̄ =
1
n2
n∑i=1
σ2Xi
=1
n2
n∑i=1
σ2 =σ2
n
so, σx̄ = σ/√n
March 12, 2015 8 / 28
![Page 29: Chapter 5: Sampling DistributionsChapter 5: Sampling Distributions Tuesday (3/10/15): 5.1 {Sampling Distributions {Sample means and the Law of Large Numbers {The Central Limit Theorem](https://reader030.fdocuments.us/reader030/viewer/2022040223/5e5413034285d826427cbc6a/html5/thumbnails/29.jpg)
Mean and standard deviation of the sample mean
– Knowing that µXi = µ for all i, we can compute the mean of x̄ (recallthat x̄ is a random variable!)
µx̄ = E[x̄] = E
[1
n
n∑i=1
Xi
]=
1
n
[n∑
i=1
E[Xi]
]=
1
n
n∑i=1
µ =1
nn · µ = µ
So, this says that x̄ is an unbiased estimator of µ, since the mean ofthe sampling distribution is the same as the mean of the population!– Note: the assumption that n is small compared to the population iscrucial. If n is not small compared to population, Law of LargeNumbers is more useful.– We can also compute the sample variance/standard deviation:
σ2x̄ =
1
n2
n∑i=1
σ2Xi
=1
n2
n∑i=1
σ2 =σ2
n
so, σx̄ = σ/√n
March 12, 2015 8 / 28
![Page 30: Chapter 5: Sampling DistributionsChapter 5: Sampling Distributions Tuesday (3/10/15): 5.1 {Sampling Distributions {Sample means and the Law of Large Numbers {The Central Limit Theorem](https://reader030.fdocuments.us/reader030/viewer/2022040223/5e5413034285d826427cbc6a/html5/thumbnails/30.jpg)
Interpretation
– Recall the setting: we are taking simple random samples of size nfrom a population of size much bigger than n.
– We can estimate the population mean by computing x̄. This is anunbiased estimate because if we took x̄ for every possible sample ofsize n, the distribution of x̄ would be centered around µ.
– The larger we take n (but not too big...), the smaller σx̄ gets. Inother words, the distribution of x̄ becomes less variable if wetake larger samples.
– Intuition: averaging (computing x̄ over a sample smooths out thevariability, making sample means less variable than individualobservations.
– The question still remains:
What does the distribution of x̄ look like?
March 12, 2015 9 / 28
![Page 31: Chapter 5: Sampling DistributionsChapter 5: Sampling Distributions Tuesday (3/10/15): 5.1 {Sampling Distributions {Sample means and the Law of Large Numbers {The Central Limit Theorem](https://reader030.fdocuments.us/reader030/viewer/2022040223/5e5413034285d826427cbc6a/html5/thumbnails/31.jpg)
Interpretation
– Recall the setting: we are taking simple random samples of size nfrom a population of size much bigger than n.
– We can estimate the population mean by computing x̄. This is anunbiased estimate because if we took x̄ for every possible sample ofsize n, the distribution of x̄ would be centered around µ.
– The larger we take n (but not too big...), the smaller σx̄ gets. Inother words, the distribution of x̄ becomes less variable if wetake larger samples.
– Intuition: averaging (computing x̄ over a sample smooths out thevariability, making sample means less variable than individualobservations.
– The question still remains:
What does the distribution of x̄ look like?
March 12, 2015 9 / 28
![Page 32: Chapter 5: Sampling DistributionsChapter 5: Sampling Distributions Tuesday (3/10/15): 5.1 {Sampling Distributions {Sample means and the Law of Large Numbers {The Central Limit Theorem](https://reader030.fdocuments.us/reader030/viewer/2022040223/5e5413034285d826427cbc6a/html5/thumbnails/32.jpg)
Interpretation
– Recall the setting: we are taking simple random samples of size nfrom a population of size much bigger than n.
– We can estimate the population mean by computing x̄. This is anunbiased estimate because if we took x̄ for every possible sample ofsize n, the distribution of x̄ would be centered around µ.
– The larger we take n (but not too big...), the smaller σx̄ gets. Inother words, the distribution of x̄ becomes less variable if wetake larger samples.
– Intuition: averaging (computing x̄ over a sample smooths out thevariability, making sample means less variable than individualobservations.
– The question still remains:
What does the distribution of x̄ look like?
March 12, 2015 9 / 28
![Page 33: Chapter 5: Sampling DistributionsChapter 5: Sampling Distributions Tuesday (3/10/15): 5.1 {Sampling Distributions {Sample means and the Law of Large Numbers {The Central Limit Theorem](https://reader030.fdocuments.us/reader030/viewer/2022040223/5e5413034285d826427cbc6a/html5/thumbnails/33.jpg)
Interpretation
– Recall the setting: we are taking simple random samples of size nfrom a population of size much bigger than n.
– We can estimate the population mean by computing x̄. This is anunbiased estimate because if we took x̄ for every possible sample ofsize n, the distribution of x̄ would be centered around µ.
– The larger we take n (but not too big...), the smaller σx̄ gets. Inother words, the distribution of x̄ becomes less variable if wetake larger samples.
– Intuition: averaging (computing x̄ over a sample smooths out thevariability, making sample means less variable than individualobservations.
– The question still remains:
What does the distribution of x̄ look like?
March 12, 2015 9 / 28
![Page 34: Chapter 5: Sampling DistributionsChapter 5: Sampling Distributions Tuesday (3/10/15): 5.1 {Sampling Distributions {Sample means and the Law of Large Numbers {The Central Limit Theorem](https://reader030.fdocuments.us/reader030/viewer/2022040223/5e5413034285d826427cbc6a/html5/thumbnails/34.jpg)
Interpretation
– Recall the setting: we are taking simple random samples of size nfrom a population of size much bigger than n.
– We can estimate the population mean by computing x̄. This is anunbiased estimate because if we took x̄ for every possible sample ofsize n, the distribution of x̄ would be centered around µ.
– The larger we take n (but not too big...), the smaller σx̄ gets. Inother words, the distribution of x̄ becomes less variable if wetake larger samples.
– Intuition: averaging (computing x̄ over a sample smooths out thevariability, making sample means less variable than individualobservations.
– The question still remains:
What does the distribution of x̄ look like?
March 12, 2015 9 / 28
![Page 35: Chapter 5: Sampling DistributionsChapter 5: Sampling Distributions Tuesday (3/10/15): 5.1 {Sampling Distributions {Sample means and the Law of Large Numbers {The Central Limit Theorem](https://reader030.fdocuments.us/reader030/viewer/2022040223/5e5413034285d826427cbc6a/html5/thumbnails/35.jpg)
The distribution of x̄
– Using excel, we generate uniform random whole numbers from{1, 2, 3, 4, 5} using RANDBETWEEN(1,5). This is like drawingrandom people from our UA student example - recall that each of50,000 people had a number between 1 and 5, with the same numberof people holding each number.
– To simulate random sampling of size n, we can simply use a range ofcells of size n, i.e. A1:A5 or A11:A15 are simple random samples ofsize 5.
– We compute x̄ for different random samples of size n, then make ahistogram of x̄
March 12, 2015 10 / 28
![Page 36: Chapter 5: Sampling DistributionsChapter 5: Sampling Distributions Tuesday (3/10/15): 5.1 {Sampling Distributions {Sample means and the Law of Large Numbers {The Central Limit Theorem](https://reader030.fdocuments.us/reader030/viewer/2022040223/5e5413034285d826427cbc6a/html5/thumbnails/36.jpg)
The distribution of x̄
– Using excel, we generate uniform random whole numbers from{1, 2, 3, 4, 5} using RANDBETWEEN(1,5). This is like drawingrandom people from our UA student example - recall that each of50,000 people had a number between 1 and 5, with the same numberof people holding each number.
– To simulate random sampling of size n, we can simply use a range ofcells of size n, i.e. A1:A5 or A11:A15 are simple random samples ofsize 5.
– We compute x̄ for different random samples of size n, then make ahistogram of x̄
March 12, 2015 10 / 28
![Page 37: Chapter 5: Sampling DistributionsChapter 5: Sampling Distributions Tuesday (3/10/15): 5.1 {Sampling Distributions {Sample means and the Law of Large Numbers {The Central Limit Theorem](https://reader030.fdocuments.us/reader030/viewer/2022040223/5e5413034285d826427cbc6a/html5/thumbnails/37.jpg)
The distribution of x̄
– Using excel, we generate uniform random whole numbers from{1, 2, 3, 4, 5} using RANDBETWEEN(1,5). This is like drawingrandom people from our UA student example - recall that each of50,000 people had a number between 1 and 5, with the same numberof people holding each number.
– To simulate random sampling of size n, we can simply use a range ofcells of size n, i.e. A1:A5 or A11:A15 are simple random samples ofsize 5.
– We compute x̄ for different random samples of size n, then make ahistogram of x̄
March 12, 2015 10 / 28
![Page 38: Chapter 5: Sampling DistributionsChapter 5: Sampling Distributions Tuesday (3/10/15): 5.1 {Sampling Distributions {Sample means and the Law of Large Numbers {The Central Limit Theorem](https://reader030.fdocuments.us/reader030/viewer/2022040223/5e5413034285d826427cbc6a/html5/thumbnails/38.jpg)
The distribution of x̄
We know that for samples of size n, x̄ has mean µ (same as thepopulation mean) and standard deviation σ/
√n. The distributions
look like:
These look suspiciously normal!
March 12, 2015 11 / 28
![Page 39: Chapter 5: Sampling DistributionsChapter 5: Sampling Distributions Tuesday (3/10/15): 5.1 {Sampling Distributions {Sample means and the Law of Large Numbers {The Central Limit Theorem](https://reader030.fdocuments.us/reader030/viewer/2022040223/5e5413034285d826427cbc6a/html5/thumbnails/39.jpg)
The distribution of x̄
We know that for samples of size n, x̄ has mean µ (same as thepopulation mean) and standard deviation σ/
√n. The distributions
look like:
These look suspiciously normal!
March 12, 2015 11 / 28
![Page 40: Chapter 5: Sampling DistributionsChapter 5: Sampling Distributions Tuesday (3/10/15): 5.1 {Sampling Distributions {Sample means and the Law of Large Numbers {The Central Limit Theorem](https://reader030.fdocuments.us/reader030/viewer/2022040223/5e5413034285d826427cbc6a/html5/thumbnails/40.jpg)
The distribution of x̄
We know that for samples of size n, x̄ has mean µ (same as thepopulation mean) and standard deviation σ/
√n. The distributions
look like:
These look suspiciously normal!
March 12, 2015 11 / 28
![Page 41: Chapter 5: Sampling DistributionsChapter 5: Sampling Distributions Tuesday (3/10/15): 5.1 {Sampling Distributions {Sample means and the Law of Large Numbers {The Central Limit Theorem](https://reader030.fdocuments.us/reader030/viewer/2022040223/5e5413034285d826427cbc6a/html5/thumbnails/41.jpg)
The distribution of x̄
We know that for samples of size n, x̄ has mean µ (same as thepopulation mean) and standard deviation σ/
√n. The distributions
look like:
These look suspiciously normal!
March 12, 2015 11 / 28
![Page 42: Chapter 5: Sampling DistributionsChapter 5: Sampling Distributions Tuesday (3/10/15): 5.1 {Sampling Distributions {Sample means and the Law of Large Numbers {The Central Limit Theorem](https://reader030.fdocuments.us/reader030/viewer/2022040223/5e5413034285d826427cbc6a/html5/thumbnails/42.jpg)
The distribution of x̄
We know that for samples of size n, x̄ has mean µ (same as thepopulation mean) and standard deviation σ/
√n. The distributions
look like:
These look suspiciously normal!
March 12, 2015 11 / 28
![Page 43: Chapter 5: Sampling DistributionsChapter 5: Sampling Distributions Tuesday (3/10/15): 5.1 {Sampling Distributions {Sample means and the Law of Large Numbers {The Central Limit Theorem](https://reader030.fdocuments.us/reader030/viewer/2022040223/5e5413034285d826427cbc6a/html5/thumbnails/43.jpg)
The distribution of x̄
We know that for samples of size n, x̄ has mean µ (same as thepopulation mean) and standard deviation σ/
√n. The distributions
look like:
These look suspiciously normal!
March 12, 2015 11 / 28
![Page 44: Chapter 5: Sampling DistributionsChapter 5: Sampling Distributions Tuesday (3/10/15): 5.1 {Sampling Distributions {Sample means and the Law of Large Numbers {The Central Limit Theorem](https://reader030.fdocuments.us/reader030/viewer/2022040223/5e5413034285d826427cbc6a/html5/thumbnails/44.jpg)
The Central Limit Theorem
Suppose that X1, X2, . . . , Xn are independent, identically distributedrandom variables. For instance, drawing random samples from a largepopulation, with n� population size, each individual shares the samedistribution as X, the population parameter. Then, the distribution of
x̄ =1
n(X1 +Xn + . . .+Xn)
is approximately normal (unless the Xi are normal) with mean µXand standard deviation σX/
√n. If the Xi are normal, then x̄ is exactly
N(µ, σ/√n).
– In other words,
x̄ ∼ N(µ, σ/√n) (Approximately)
– Note: It does not matter what the probability distribution of thevariables Xi is! This works with normally distributed RVs, uniformlydistributed RVs, and RVs with crazy skewed distributions.
March 12, 2015 12 / 28
![Page 45: Chapter 5: Sampling DistributionsChapter 5: Sampling Distributions Tuesday (3/10/15): 5.1 {Sampling Distributions {Sample means and the Law of Large Numbers {The Central Limit Theorem](https://reader030.fdocuments.us/reader030/viewer/2022040223/5e5413034285d826427cbc6a/html5/thumbnails/45.jpg)
The Central Limit Theorem
Suppose that X1, X2, . . . , Xn are independent, identically distributedrandom variables. For instance, drawing random samples from a largepopulation, with n� population size, each individual shares the samedistribution as X, the population parameter. Then, the distribution of
x̄ =1
n(X1 +Xn + . . .+Xn)
is approximately normal (unless the Xi are normal) with mean µXand standard deviation σX/
√n. If the Xi are normal, then x̄ is exactly
N(µ, σ/√n).
– In other words,
x̄ ∼ N(µ, σ/√n) (Approximately)
– Note: It does not matter what the probability distribution of thevariables Xi is! This works with normally distributed RVs, uniformlydistributed RVs, and RVs with crazy skewed distributions.
March 12, 2015 12 / 28
![Page 46: Chapter 5: Sampling DistributionsChapter 5: Sampling Distributions Tuesday (3/10/15): 5.1 {Sampling Distributions {Sample means and the Law of Large Numbers {The Central Limit Theorem](https://reader030.fdocuments.us/reader030/viewer/2022040223/5e5413034285d826427cbc6a/html5/thumbnails/46.jpg)
The Central Limit Theorem
Suppose that X1, X2, . . . , Xn are independent, identically distributedrandom variables. For instance, drawing random samples from a largepopulation, with n� population size, each individual shares the samedistribution as X, the population parameter. Then, the distribution of
x̄ =1
n(X1 +Xn + . . .+Xn)
is approximately normal (unless the Xi are normal) with mean µXand standard deviation σX/
√n. If the Xi are normal, then x̄ is exactly
N(µ, σ/√n).
– In other words,
x̄ ∼ N(µ, σ/√n) (Approximately)
– Note: It does not matter what the probability distribution of thevariables Xi is! This works with normally distributed RVs, uniformlydistributed RVs, and RVs with crazy skewed distributions.
March 12, 2015 12 / 28
![Page 47: Chapter 5: Sampling DistributionsChapter 5: Sampling Distributions Tuesday (3/10/15): 5.1 {Sampling Distributions {Sample means and the Law of Large Numbers {The Central Limit Theorem](https://reader030.fdocuments.us/reader030/viewer/2022040223/5e5413034285d826427cbc6a/html5/thumbnails/47.jpg)
Example
Suppose that we draw random real numbers uniformly from 0 to 10:
0.1 Mean = 5, standard deviation = 2.88
a) What is the probability that a given number drawn is between 1and 2?
b) Suppose we select 25 numbers randomly. Approximately what is theprobability that the mean of these numbers is between 1 and 2?
Answer :
a) Find the area under the PDF between 1 and 2.
b) Using the CLT, we know that x̄ ∼ N(5, 2.88/√
25), so we use
normalcdf(1,2,5,2.88/5)
March 12, 2015 13 / 28
![Page 48: Chapter 5: Sampling DistributionsChapter 5: Sampling Distributions Tuesday (3/10/15): 5.1 {Sampling Distributions {Sample means and the Law of Large Numbers {The Central Limit Theorem](https://reader030.fdocuments.us/reader030/viewer/2022040223/5e5413034285d826427cbc6a/html5/thumbnails/48.jpg)
Example
Suppose that we draw random real numbers uniformly from 0 to 10:
0.1 Mean = 5, standard deviation = 2.88
a) What is the probability that a given number drawn is between 1and 2?
b) Suppose we select 25 numbers randomly. Approximately what is theprobability that the mean of these numbers is between 1 and 2?
Answer :
a) Find the area under the PDF between 1 and 2.
b) Using the CLT, we know that x̄ ∼ N(5, 2.88/√
25), so we use
normalcdf(1,2,5,2.88/5)
March 12, 2015 13 / 28
![Page 49: Chapter 5: Sampling DistributionsChapter 5: Sampling Distributions Tuesday (3/10/15): 5.1 {Sampling Distributions {Sample means and the Law of Large Numbers {The Central Limit Theorem](https://reader030.fdocuments.us/reader030/viewer/2022040223/5e5413034285d826427cbc6a/html5/thumbnails/49.jpg)
Example
Suppose ACT scores across all incoming freshmen in the US arenormally distributed with a mean of 20.8 and a standard deviation of4.8.
a) Are the given means and standard deviations for the population orfor a sample?
b) What is the probability that a randomly selected student had ascore higher than 30?
c) What is the probability that 5 randomly selected students havetheir average score higher than 30?
Answers:
a) These are for the population
b) Since this random variable is normally distributed, we can usenormalcdf(30,1000,20.8,4.8) = 0.028 (2 sig figs). As usual, the1000 is just a ‘large’ number to indicate that we want all numbersto the right of 30.
March 12, 2015 14 / 28
![Page 50: Chapter 5: Sampling DistributionsChapter 5: Sampling Distributions Tuesday (3/10/15): 5.1 {Sampling Distributions {Sample means and the Law of Large Numbers {The Central Limit Theorem](https://reader030.fdocuments.us/reader030/viewer/2022040223/5e5413034285d826427cbc6a/html5/thumbnails/50.jpg)
Example
Suppose ACT scores across all incoming freshmen in the US arenormally distributed with a mean of 20.8 and a standard deviation of4.8.
a) Are the given means and standard deviations for the population orfor a sample?
b) What is the probability that a randomly selected student had ascore higher than 30?
c) What is the probability that 5 randomly selected students havetheir average score higher than 30?
Answers:
a) These are for the population
b) Since this random variable is normally distributed, we can usenormalcdf(30,1000,20.8,4.8) = 0.028 (2 sig figs). As usual, the1000 is just a ‘large’ number to indicate that we want all numbersto the right of 30.
March 12, 2015 14 / 28
![Page 51: Chapter 5: Sampling DistributionsChapter 5: Sampling Distributions Tuesday (3/10/15): 5.1 {Sampling Distributions {Sample means and the Law of Large Numbers {The Central Limit Theorem](https://reader030.fdocuments.us/reader030/viewer/2022040223/5e5413034285d826427cbc6a/html5/thumbnails/51.jpg)
Example, cont’d
Solution to part c): by the central limit theorem, we know that
x̄ ∼ N(µ,
σ√5
)Since our original random variable is normal, this is in fact exact - x̄will have exactly a normal distribution with mean µ = 20.8 andstandard deviation σ/
√5 (5 is the sample size). Thus the probability
that 5 randomly selected students has an average score higher than 30
is found with normalcdf(30,1000,20.8,4.8/√
5) = 9.11 · 10−6
March 12, 2015 15 / 28
![Page 52: Chapter 5: Sampling DistributionsChapter 5: Sampling Distributions Tuesday (3/10/15): 5.1 {Sampling Distributions {Sample means and the Law of Large Numbers {The Central Limit Theorem](https://reader030.fdocuments.us/reader030/viewer/2022040223/5e5413034285d826427cbc6a/html5/thumbnails/52.jpg)
Review of LLN and CLT
Assumptions:
1) I have n independent identically distributed random variablesX1, . . . , Xn, each with mean µ and standard deviation σ.
In most examples, the situation is this: each Xi represents avariable for an individual drawn from a very large population, sothat the collection {X1, . . . , Xn} models a sample of size n. Forexample, each Xi might be a randomly selected person’s age or
height or blood O2 level.
2) The random variables must have finite mean and standarddeviation (not an issue in this class, but it does come up!)
Define the sample mean as the random variable
X̄ =1
n(X1 +X2 + . . .+Xn) =
1
n
n∑i=1
Xi
March 12, 2015 16 / 28
![Page 53: Chapter 5: Sampling DistributionsChapter 5: Sampling Distributions Tuesday (3/10/15): 5.1 {Sampling Distributions {Sample means and the Law of Large Numbers {The Central Limit Theorem](https://reader030.fdocuments.us/reader030/viewer/2022040223/5e5413034285d826427cbc6a/html5/thumbnails/53.jpg)
Review of LLN and CLT
Assumptions:
1) I have n independent identically distributed random variablesX1, . . . , Xn, each with mean µ and standard deviation σ.
In most examples, the situation is this: each Xi represents avariable for an individual drawn from a very large population, sothat the collection {X1, . . . , Xn} models a sample of size n. Forexample, each Xi might be a randomly selected person’s age or
height or blood O2 level.
2) The random variables must have finite mean and standarddeviation (not an issue in this class, but it does come up!)
Define the sample mean as the random variable
X̄ =1
n(X1 +X2 + . . .+Xn) =
1
n
n∑i=1
Xi
March 12, 2015 16 / 28
![Page 54: Chapter 5: Sampling DistributionsChapter 5: Sampling Distributions Tuesday (3/10/15): 5.1 {Sampling Distributions {Sample means and the Law of Large Numbers {The Central Limit Theorem](https://reader030.fdocuments.us/reader030/viewer/2022040223/5e5413034285d826427cbc6a/html5/thumbnails/54.jpg)
Review of LLN and CLT
Assumptions:
1) I have n independent identically distributed random variablesX1, . . . , Xn, each with mean µ and standard deviation σ.
In most examples, the situation is this: each Xi represents avariable for an individual drawn from a very large population, sothat the collection {X1, . . . , Xn} models a sample of size n. Forexample, each Xi might be a randomly selected person’s age or
height or blood O2 level.
2) The random variables must have finite mean and standarddeviation (not an issue in this class, but it does come up!)
Define the sample mean as the random variable
X̄ =1
n(X1 +X2 + . . .+Xn) =
1
n
n∑i=1
Xi
March 12, 2015 16 / 28
![Page 55: Chapter 5: Sampling DistributionsChapter 5: Sampling Distributions Tuesday (3/10/15): 5.1 {Sampling Distributions {Sample means and the Law of Large Numbers {The Central Limit Theorem](https://reader030.fdocuments.us/reader030/viewer/2022040223/5e5413034285d826427cbc6a/html5/thumbnails/55.jpg)
Review of LLN and CLT
Conclusions:
1) Law of large numbers: X̄ → µ. Loosely speaking we mean that aswe collect larger and larger samples, the sample mean ‘goes to’ thepopulation mean.
2) Central limit theorem: the variation of X̄ around µ follows anormal distribution. More precisely,
X̄ ≈ N(µ,
σ√n
)– This means that if we collected many samples of a fixed size n, wewould expect the sample means of these samples to be clusteredaround the actual population mean. The variability decreases withlarger samples, as long as we still satisfy the IID assumption.
March 12, 2015 17 / 28
![Page 56: Chapter 5: Sampling DistributionsChapter 5: Sampling Distributions Tuesday (3/10/15): 5.1 {Sampling Distributions {Sample means and the Law of Large Numbers {The Central Limit Theorem](https://reader030.fdocuments.us/reader030/viewer/2022040223/5e5413034285d826427cbc6a/html5/thumbnails/56.jpg)
5.2: Repeated Trials and the Binomial Distribution
– A Bernoulli trial is an experiment where the outcome is one of twopossibilities - i.e. ‘yes’ or ‘no’, ‘heads’ or ‘tails’, etc.
– When we repeat Bernoulli trials and record the number of positiveresults, we are in a Binomial setting.
More precisely, we need:
– A binary experiment (outcome is one of two options)
– Each ‘trial’ is independent
– We fix a number of trials at the start
– Each trial has the same probability p of success and probability 1− pof failure.
Example: flip a fair coin 5 times and count the number of heads. Whatis p? Answer: p = 1/2 since the coin is fair.
March 12, 2015 18 / 28
![Page 57: Chapter 5: Sampling DistributionsChapter 5: Sampling Distributions Tuesday (3/10/15): 5.1 {Sampling Distributions {Sample means and the Law of Large Numbers {The Central Limit Theorem](https://reader030.fdocuments.us/reader030/viewer/2022040223/5e5413034285d826427cbc6a/html5/thumbnails/57.jpg)
5.2: Repeated Trials and the Binomial Distribution
– A Bernoulli trial is an experiment where the outcome is one of twopossibilities - i.e. ‘yes’ or ‘no’, ‘heads’ or ‘tails’, etc.
– When we repeat Bernoulli trials and record the number of positiveresults, we are in a Binomial setting.
More precisely, we need:
– A binary experiment (outcome is one of two options)
– Each ‘trial’ is independent
– We fix a number of trials at the start
– Each trial has the same probability p of success and probability 1− pof failure.
Example: flip a fair coin 5 times and count the number of heads. Whatis p? Answer: p = 1/2 since the coin is fair.
March 12, 2015 18 / 28
![Page 58: Chapter 5: Sampling DistributionsChapter 5: Sampling Distributions Tuesday (3/10/15): 5.1 {Sampling Distributions {Sample means and the Law of Large Numbers {The Central Limit Theorem](https://reader030.fdocuments.us/reader030/viewer/2022040223/5e5413034285d826427cbc6a/html5/thumbnails/58.jpg)
5.2: Repeated Trials and the Binomial Distribution
– A Bernoulli trial is an experiment where the outcome is one of twopossibilities - i.e. ‘yes’ or ‘no’, ‘heads’ or ‘tails’, etc.
– When we repeat Bernoulli trials and record the number of positiveresults, we are in a Binomial setting.
More precisely, we need:
– A binary experiment (outcome is one of two options)
– Each ‘trial’ is independent
– We fix a number of trials at the start
– Each trial has the same probability p of success and probability 1− pof failure.
Example: flip a fair coin 5 times and count the number of heads. Whatis p?
Answer: p = 1/2 since the coin is fair.
March 12, 2015 18 / 28
![Page 59: Chapter 5: Sampling DistributionsChapter 5: Sampling Distributions Tuesday (3/10/15): 5.1 {Sampling Distributions {Sample means and the Law of Large Numbers {The Central Limit Theorem](https://reader030.fdocuments.us/reader030/viewer/2022040223/5e5413034285d826427cbc6a/html5/thumbnails/59.jpg)
5.2: Repeated Trials and the Binomial Distribution
– A Bernoulli trial is an experiment where the outcome is one of twopossibilities - i.e. ‘yes’ or ‘no’, ‘heads’ or ‘tails’, etc.
– When we repeat Bernoulli trials and record the number of positiveresults, we are in a Binomial setting.
More precisely, we need:
– A binary experiment (outcome is one of two options)
– Each ‘trial’ is independent
– We fix a number of trials at the start
– Each trial has the same probability p of success and probability 1− pof failure.
Example: flip a fair coin 5 times and count the number of heads. Whatis p? Answer: p = 1/2 since the coin is fair.
March 12, 2015 18 / 28
![Page 60: Chapter 5: Sampling DistributionsChapter 5: Sampling Distributions Tuesday (3/10/15): 5.1 {Sampling Distributions {Sample means and the Law of Large Numbers {The Central Limit Theorem](https://reader030.fdocuments.us/reader030/viewer/2022040223/5e5413034285d826427cbc6a/html5/thumbnails/60.jpg)
Example
Suppose we roll a (fair) 6-sided die. A ‘success’ (S) will be if the dieshows a 6. Any other number is considered a fail (F) We will roll thedie 3 times and count the number of successes.
– What is the probability of a success on a given trial? Answer:p = 1/6.
– What is the probability of the following: SFF? Answer:Pr(SFF ) = (1/6) · (5/6) · (5/6) = 25
216 ≈ 0.116
– What is the probability of getting exactly one success in 3 rolls?Answer: there are 3 different ways to get one success:SFF, FSF, FFS. Each has the same probability, 25/216. So, theprobability of getting exactly one success is
325
216=
75
216≈ 0.347
March 12, 2015 19 / 28
![Page 61: Chapter 5: Sampling DistributionsChapter 5: Sampling Distributions Tuesday (3/10/15): 5.1 {Sampling Distributions {Sample means and the Law of Large Numbers {The Central Limit Theorem](https://reader030.fdocuments.us/reader030/viewer/2022040223/5e5413034285d826427cbc6a/html5/thumbnails/61.jpg)
Example
Suppose we roll a (fair) 6-sided die. A ‘success’ (S) will be if the dieshows a 6. Any other number is considered a fail (F) We will roll thedie 3 times and count the number of successes.
– What is the probability of a success on a given trial?
Answer:p = 1/6.
– What is the probability of the following: SFF? Answer:Pr(SFF ) = (1/6) · (5/6) · (5/6) = 25
216 ≈ 0.116
– What is the probability of getting exactly one success in 3 rolls?Answer: there are 3 different ways to get one success:SFF, FSF, FFS. Each has the same probability, 25/216. So, theprobability of getting exactly one success is
325
216=
75
216≈ 0.347
March 12, 2015 19 / 28
![Page 62: Chapter 5: Sampling DistributionsChapter 5: Sampling Distributions Tuesday (3/10/15): 5.1 {Sampling Distributions {Sample means and the Law of Large Numbers {The Central Limit Theorem](https://reader030.fdocuments.us/reader030/viewer/2022040223/5e5413034285d826427cbc6a/html5/thumbnails/62.jpg)
Example
Suppose we roll a (fair) 6-sided die. A ‘success’ (S) will be if the dieshows a 6. Any other number is considered a fail (F) We will roll thedie 3 times and count the number of successes.
– What is the probability of a success on a given trial? Answer:p = 1/6.
– What is the probability of the following: SFF? Answer:Pr(SFF ) = (1/6) · (5/6) · (5/6) = 25
216 ≈ 0.116
– What is the probability of getting exactly one success in 3 rolls?Answer: there are 3 different ways to get one success:SFF, FSF, FFS. Each has the same probability, 25/216. So, theprobability of getting exactly one success is
325
216=
75
216≈ 0.347
March 12, 2015 19 / 28
![Page 63: Chapter 5: Sampling DistributionsChapter 5: Sampling Distributions Tuesday (3/10/15): 5.1 {Sampling Distributions {Sample means and the Law of Large Numbers {The Central Limit Theorem](https://reader030.fdocuments.us/reader030/viewer/2022040223/5e5413034285d826427cbc6a/html5/thumbnails/63.jpg)
Example
Suppose we roll a (fair) 6-sided die. A ‘success’ (S) will be if the dieshows a 6. Any other number is considered a fail (F) We will roll thedie 3 times and count the number of successes.
– What is the probability of a success on a given trial? Answer:p = 1/6.
– What is the probability of the following: SFF?
Answer:Pr(SFF ) = (1/6) · (5/6) · (5/6) = 25
216 ≈ 0.116
– What is the probability of getting exactly one success in 3 rolls?Answer: there are 3 different ways to get one success:SFF, FSF, FFS. Each has the same probability, 25/216. So, theprobability of getting exactly one success is
325
216=
75
216≈ 0.347
March 12, 2015 19 / 28
![Page 64: Chapter 5: Sampling DistributionsChapter 5: Sampling Distributions Tuesday (3/10/15): 5.1 {Sampling Distributions {Sample means and the Law of Large Numbers {The Central Limit Theorem](https://reader030.fdocuments.us/reader030/viewer/2022040223/5e5413034285d826427cbc6a/html5/thumbnails/64.jpg)
Example
Suppose we roll a (fair) 6-sided die. A ‘success’ (S) will be if the dieshows a 6. Any other number is considered a fail (F) We will roll thedie 3 times and count the number of successes.
– What is the probability of a success on a given trial? Answer:p = 1/6.
– What is the probability of the following: SFF? Answer:Pr(SFF ) = (1/6) · (5/6) · (5/6) = 25
216 ≈ 0.116
– What is the probability of getting exactly one success in 3 rolls?Answer: there are 3 different ways to get one success:SFF, FSF, FFS. Each has the same probability, 25/216. So, theprobability of getting exactly one success is
325
216=
75
216≈ 0.347
March 12, 2015 19 / 28
![Page 65: Chapter 5: Sampling DistributionsChapter 5: Sampling Distributions Tuesday (3/10/15): 5.1 {Sampling Distributions {Sample means and the Law of Large Numbers {The Central Limit Theorem](https://reader030.fdocuments.us/reader030/viewer/2022040223/5e5413034285d826427cbc6a/html5/thumbnails/65.jpg)
Example
Suppose we roll a (fair) 6-sided die. A ‘success’ (S) will be if the dieshows a 6. Any other number is considered a fail (F) We will roll thedie 3 times and count the number of successes.
– What is the probability of a success on a given trial? Answer:p = 1/6.
– What is the probability of the following: SFF? Answer:Pr(SFF ) = (1/6) · (5/6) · (5/6) = 25
216 ≈ 0.116
– What is the probability of getting exactly one success in 3 rolls?
Answer: there are 3 different ways to get one success:SFF, FSF, FFS. Each has the same probability, 25/216. So, theprobability of getting exactly one success is
325
216=
75
216≈ 0.347
March 12, 2015 19 / 28
![Page 66: Chapter 5: Sampling DistributionsChapter 5: Sampling Distributions Tuesday (3/10/15): 5.1 {Sampling Distributions {Sample means and the Law of Large Numbers {The Central Limit Theorem](https://reader030.fdocuments.us/reader030/viewer/2022040223/5e5413034285d826427cbc6a/html5/thumbnails/66.jpg)
Example
Suppose we roll a (fair) 6-sided die. A ‘success’ (S) will be if the dieshows a 6. Any other number is considered a fail (F) We will roll thedie 3 times and count the number of successes.
– What is the probability of a success on a given trial? Answer:p = 1/6.
– What is the probability of the following: SFF? Answer:Pr(SFF ) = (1/6) · (5/6) · (5/6) = 25
216 ≈ 0.116
– What is the probability of getting exactly one success in 3 rolls?Answer: there are 3 different ways to get one success:SFF, FSF, FFS. Each has the same probability, 25/216. So, theprobability of getting exactly one success is
325
216=
75
216≈ 0.347
March 12, 2015 19 / 28
![Page 67: Chapter 5: Sampling DistributionsChapter 5: Sampling Distributions Tuesday (3/10/15): 5.1 {Sampling Distributions {Sample means and the Law of Large Numbers {The Central Limit Theorem](https://reader030.fdocuments.us/reader030/viewer/2022040223/5e5413034285d826427cbc6a/html5/thumbnails/67.jpg)
Example
Suppose flip an unfair coin 100 times. The probability of heads is 0.4.What is the probability that we get exactly 20 heads in 100 flips?
Solution: Let X be the number of heads in 100 flips. The probabilityof getting exactly 20 heads will be
P (X = 20) = N0.4200.680
where N is the number of different ways to get 20 heads. We can findN using the ‘choose’ function:(
100
20
)(‘100 choose 20’)
March 12, 2015 20 / 28
![Page 68: Chapter 5: Sampling DistributionsChapter 5: Sampling Distributions Tuesday (3/10/15): 5.1 {Sampling Distributions {Sample means and the Law of Large Numbers {The Central Limit Theorem](https://reader030.fdocuments.us/reader030/viewer/2022040223/5e5413034285d826427cbc6a/html5/thumbnails/68.jpg)
Example
Suppose flip an unfair coin 100 times. The probability of heads is 0.4.What is the probability that we get exactly 20 heads in 100 flips?Solution: Let X be the number of heads in 100 flips. The probabilityof getting exactly 20 heads will be
P (X = 20) = N0.4200.680
where N is the number of different ways to get 20 heads. We can findN using the ‘choose’ function:(
100
20
)(‘100 choose 20’)
March 12, 2015 20 / 28
![Page 69: Chapter 5: Sampling DistributionsChapter 5: Sampling Distributions Tuesday (3/10/15): 5.1 {Sampling Distributions {Sample means and the Law of Large Numbers {The Central Limit Theorem](https://reader030.fdocuments.us/reader030/viewer/2022040223/5e5413034285d826427cbc6a/html5/thumbnails/69.jpg)
The Binomial Distribution
When we have repeated Bernoulli trials that satisfy the conditions fora binomial setting (independence, fixed number of trials n, fixedprobability of success p), the distribution for the number of success isthe Binomial Distribution:
P (X = k) =
(n
k
)pk(1− p)n−k
Intuitively,
– pk is the probability of getting k successes– (1− p)n−k is the probability of the rest (n− k) being failures– pk(1− p)n−k is the probability of getting exactly k success and n− k
failures–(nk
)is the number of different ways this can happen.
The number(nk
)is the binomial coefficient:(n
k
)=
n!
k!(n− k)!, n! = n · (n− 1) · · · 2 · 1
March 12, 2015 21 / 28
![Page 70: Chapter 5: Sampling DistributionsChapter 5: Sampling Distributions Tuesday (3/10/15): 5.1 {Sampling Distributions {Sample means and the Law of Large Numbers {The Central Limit Theorem](https://reader030.fdocuments.us/reader030/viewer/2022040223/5e5413034285d826427cbc6a/html5/thumbnails/70.jpg)
The Binomial Distribution
When we have repeated Bernoulli trials that satisfy the conditions fora binomial setting (independence, fixed number of trials n, fixedprobability of success p), the distribution for the number of success isthe Binomial Distribution:
P (X = k) =
(n
k
)pk(1− p)n−k
Intuitively,
– pk is the probability of getting k successes– (1− p)n−k is the probability of the rest (n− k) being failures– pk(1− p)n−k is the probability of getting exactly k success and n− k
failures–(nk
)is the number of different ways this can happen.
The number(nk
)is the binomial coefficient:(n
k
)=
n!
k!(n− k)!, n! = n · (n− 1) · · · 2 · 1
March 12, 2015 21 / 28
![Page 71: Chapter 5: Sampling DistributionsChapter 5: Sampling Distributions Tuesday (3/10/15): 5.1 {Sampling Distributions {Sample means and the Law of Large Numbers {The Central Limit Theorem](https://reader030.fdocuments.us/reader030/viewer/2022040223/5e5413034285d826427cbc6a/html5/thumbnails/71.jpg)
The Binomial Distribution
When we have repeated Bernoulli trials that satisfy the conditions fora binomial setting (independence, fixed number of trials n, fixedprobability of success p), the distribution for the number of success isthe Binomial Distribution:
P (X = k) =
(n
k
)pk(1− p)n−k
Intuitively,
– pk is the probability of getting k successes– (1− p)n−k is the probability of the rest (n− k) being failures– pk(1− p)n−k is the probability of getting exactly k success and n− k
failures–(nk
)is the number of different ways this can happen.
The number(nk
)is the binomial coefficient:(n
k
)=
n!
k!(n− k)!, n! = n · (n− 1) · · · 2 · 1
March 12, 2015 21 / 28
![Page 72: Chapter 5: Sampling DistributionsChapter 5: Sampling Distributions Tuesday (3/10/15): 5.1 {Sampling Distributions {Sample means and the Law of Large Numbers {The Central Limit Theorem](https://reader030.fdocuments.us/reader030/viewer/2022040223/5e5413034285d826427cbc6a/html5/thumbnails/72.jpg)
Computing Binomial Coefficients
– On the TI-84, you can compute(nk
)by inputting n, then selecting
MATH:PRB:nCr, then inputting k. It will look like this:
5 nCr 3
– You can compute binomial coefficients by hand using Pascal’striangle:
March 12, 2015 22 / 28
![Page 73: Chapter 5: Sampling DistributionsChapter 5: Sampling Distributions Tuesday (3/10/15): 5.1 {Sampling Distributions {Sample means and the Law of Large Numbers {The Central Limit Theorem](https://reader030.fdocuments.us/reader030/viewer/2022040223/5e5413034285d826427cbc6a/html5/thumbnails/73.jpg)
Computing Binomial Coefficients
– On the TI-84, you can compute(nk
)by inputting n, then selecting
MATH:PRB:nCr, then inputting k. It will look like this:
5 nCr 3
– You can compute binomial coefficients by hand using Pascal’striangle:
March 12, 2015 22 / 28
![Page 74: Chapter 5: Sampling DistributionsChapter 5: Sampling Distributions Tuesday (3/10/15): 5.1 {Sampling Distributions {Sample means and the Law of Large Numbers {The Central Limit Theorem](https://reader030.fdocuments.us/reader030/viewer/2022040223/5e5413034285d826427cbc6a/html5/thumbnails/74.jpg)
Example of Binomial Distribution?
– Example: suppose we are going to test a lot of 1000 climbing ropesfor strength. We select 5 ropes at random and test them. Supposethe probability of a rope passing a test is 0.999 (i.e. 999/1000 ropeswill pass). Define the random variable X to be the number of ropesthat fail the test in our sample.
– Is this an example of a binomial setting?
– Answer: No, not quite - each trial is not independent since thepopulation gets smaller each time.
– What is P (X = 0)?
P (X = 0) =999
1000· 998
999· 997
998· 996
997· 995
996= 0.995
– Try computing this again using a binomial distribution.
P (X = 0) ≈(
5
0
)0.0010(0.999)5 ≈ 0.99500999
March 12, 2015 23 / 28
![Page 75: Chapter 5: Sampling DistributionsChapter 5: Sampling Distributions Tuesday (3/10/15): 5.1 {Sampling Distributions {Sample means and the Law of Large Numbers {The Central Limit Theorem](https://reader030.fdocuments.us/reader030/viewer/2022040223/5e5413034285d826427cbc6a/html5/thumbnails/75.jpg)
Example of Binomial Distribution?
– Example: suppose we are going to test a lot of 1000 climbing ropesfor strength. We select 5 ropes at random and test them. Supposethe probability of a rope passing a test is 0.999 (i.e. 999/1000 ropeswill pass). Define the random variable X to be the number of ropesthat fail the test in our sample.
– Is this an example of a binomial setting?
– Answer: No, not quite - each trial is not independent since thepopulation gets smaller each time.
– What is P (X = 0)?
P (X = 0) =999
1000· 998
999· 997
998· 996
997· 995
996= 0.995
– Try computing this again using a binomial distribution.
P (X = 0) ≈(
5
0
)0.0010(0.999)5 ≈ 0.99500999
March 12, 2015 23 / 28
![Page 76: Chapter 5: Sampling DistributionsChapter 5: Sampling Distributions Tuesday (3/10/15): 5.1 {Sampling Distributions {Sample means and the Law of Large Numbers {The Central Limit Theorem](https://reader030.fdocuments.us/reader030/viewer/2022040223/5e5413034285d826427cbc6a/html5/thumbnails/76.jpg)
Example of Binomial Distribution?
– Example: suppose we are going to test a lot of 1000 climbing ropesfor strength. We select 5 ropes at random and test them. Supposethe probability of a rope passing a test is 0.999 (i.e. 999/1000 ropeswill pass). Define the random variable X to be the number of ropesthat fail the test in our sample.
– Is this an example of a binomial setting?
– Answer: No, not quite - each trial is not independent since thepopulation gets smaller each time.
– What is P (X = 0)?
P (X = 0) =999
1000· 998
999· 997
998· 996
997· 995
996= 0.995
– Try computing this again using a binomial distribution.
P (X = 0) ≈(
5
0
)0.0010(0.999)5 ≈ 0.99500999
March 12, 2015 23 / 28
![Page 77: Chapter 5: Sampling DistributionsChapter 5: Sampling Distributions Tuesday (3/10/15): 5.1 {Sampling Distributions {Sample means and the Law of Large Numbers {The Central Limit Theorem](https://reader030.fdocuments.us/reader030/viewer/2022040223/5e5413034285d826427cbc6a/html5/thumbnails/77.jpg)
Example of Binomial Distribution?
– Example: suppose we are going to test a lot of 1000 climbing ropesfor strength. We select 5 ropes at random and test them. Supposethe probability of a rope passing a test is 0.999 (i.e. 999/1000 ropeswill pass). Define the random variable X to be the number of ropesthat fail the test in our sample.
– Is this an example of a binomial setting?
– Answer: No, not quite - each trial is not independent since thepopulation gets smaller each time.
– What is P (X = 0)?
P (X = 0) =999
1000· 998
999· 997
998· 996
997· 995
996= 0.995
– Try computing this again using a binomial distribution.
P (X = 0) ≈(
5
0
)0.0010(0.999)5 ≈ 0.99500999
March 12, 2015 23 / 28
![Page 78: Chapter 5: Sampling DistributionsChapter 5: Sampling Distributions Tuesday (3/10/15): 5.1 {Sampling Distributions {Sample means and the Law of Large Numbers {The Central Limit Theorem](https://reader030.fdocuments.us/reader030/viewer/2022040223/5e5413034285d826427cbc6a/html5/thumbnails/78.jpg)
Example of Binomial Distribution?
– Example: suppose we are going to test a lot of 1000 climbing ropesfor strength. We select 5 ropes at random and test them. Supposethe probability of a rope passing a test is 0.999 (i.e. 999/1000 ropeswill pass). Define the random variable X to be the number of ropesthat fail the test in our sample.
– Is this an example of a binomial setting?
– Answer: No, not quite - each trial is not independent since thepopulation gets smaller each time.
– What is P (X = 0)?
P (X = 0) =999
1000· 998
999· 997
998· 996
997· 995
996= 0.995
– Try computing this again using a binomial distribution.
P (X = 0) ≈(
5
0
)0.0010(0.999)5 ≈ 0.99500999
March 12, 2015 23 / 28
![Page 79: Chapter 5: Sampling DistributionsChapter 5: Sampling Distributions Tuesday (3/10/15): 5.1 {Sampling Distributions {Sample means and the Law of Large Numbers {The Central Limit Theorem](https://reader030.fdocuments.us/reader030/viewer/2022040223/5e5413034285d826427cbc6a/html5/thumbnails/79.jpg)
Sampling Distribution for a ‘Count’
Suppose we sample from a population and count how many individualssatisfy a condition. For example, polling for a yes/no vote on aninitiative. We are interested in the distribution of the number of‘successes’.
If we select a simple random sample of size n from a population of sizemuch larger than n, where the probability of success is known to be p,the number of successes k in the sample is approximately binomial
with parameters p and n:
P (X = k) ≈(n
k
)pk(1− p)n−k
March 12, 2015 24 / 28
![Page 80: Chapter 5: Sampling DistributionsChapter 5: Sampling Distributions Tuesday (3/10/15): 5.1 {Sampling Distributions {Sample means and the Law of Large Numbers {The Central Limit Theorem](https://reader030.fdocuments.us/reader030/viewer/2022040223/5e5413034285d826427cbc6a/html5/thumbnails/80.jpg)
Mean and Standard Deviation of a Binomial RV
The mean and standard deviation of a binomial RV are:
µ = np
σ =√np(1− p)
Derivation is tedious, uses the binomial theorem:
(a+ b)n =
n∑k=0
(n
k
)akbn−k
March 12, 2015 25 / 28
![Page 81: Chapter 5: Sampling DistributionsChapter 5: Sampling Distributions Tuesday (3/10/15): 5.1 {Sampling Distributions {Sample means and the Law of Large Numbers {The Central Limit Theorem](https://reader030.fdocuments.us/reader030/viewer/2022040223/5e5413034285d826427cbc6a/html5/thumbnails/81.jpg)
Approximating Binomial Distributions with Normal
If the number of trials is large compared to either the probability ofsuccess or failure (say both np and n(1− p) are bigger than 10), butstill small compared to the population size, we have
Binom(n, p) ≈ N(np,√np(1− p))
Example: A survey asks a nationwide random sample of 2500 adultstheir opinion on same-sex marriage. Suppose that 55% of Americans 1
actually support same-sex marriage. Estimate the probability that1400 people or more out of the 2500 support same-sex marriage.– Check for validity of normal approximation:np = 2500 · 0.55 = 1375� 10, n(1− p) = 1125� 10.– Compute P (X ≥ 1400) using z-scores or normalcdf:
µ = np = 1375, σ =√np(1− p) ≈ 24.875
P (X ≥ 1400) ≈ normalcdf(1400, 10000, 1375, 24.875) ≈ 0.157
1see Gallup.comMarch 12, 2015 26 / 28
![Page 82: Chapter 5: Sampling DistributionsChapter 5: Sampling Distributions Tuesday (3/10/15): 5.1 {Sampling Distributions {Sample means and the Law of Large Numbers {The Central Limit Theorem](https://reader030.fdocuments.us/reader030/viewer/2022040223/5e5413034285d826427cbc6a/html5/thumbnails/82.jpg)
Approximating Binomial Distributions with Normal
If the number of trials is large compared to either the probability ofsuccess or failure (say both np and n(1− p) are bigger than 10), butstill small compared to the population size, we have
Binom(n, p) ≈ N(np,√np(1− p))
Example: A survey asks a nationwide random sample of 2500 adultstheir opinion on same-sex marriage. Suppose that 55% of Americans 1
actually support same-sex marriage. Estimate the probability that1400 people or more out of the 2500 support same-sex marriage.
– Check for validity of normal approximation:np = 2500 · 0.55 = 1375� 10, n(1− p) = 1125� 10.– Compute P (X ≥ 1400) using z-scores or normalcdf:
µ = np = 1375, σ =√np(1− p) ≈ 24.875
P (X ≥ 1400) ≈ normalcdf(1400, 10000, 1375, 24.875) ≈ 0.157
1see Gallup.comMarch 12, 2015 26 / 28
![Page 83: Chapter 5: Sampling DistributionsChapter 5: Sampling Distributions Tuesday (3/10/15): 5.1 {Sampling Distributions {Sample means and the Law of Large Numbers {The Central Limit Theorem](https://reader030.fdocuments.us/reader030/viewer/2022040223/5e5413034285d826427cbc6a/html5/thumbnails/83.jpg)
Approximating Binomial Distributions with Normal
If the number of trials is large compared to either the probability ofsuccess or failure (say both np and n(1− p) are bigger than 10), butstill small compared to the population size, we have
Binom(n, p) ≈ N(np,√np(1− p))
Example: A survey asks a nationwide random sample of 2500 adultstheir opinion on same-sex marriage. Suppose that 55% of Americans 1
actually support same-sex marriage. Estimate the probability that1400 people or more out of the 2500 support same-sex marriage.– Check for validity of normal approximation:np = 2500 · 0.55 = 1375� 10, n(1− p) = 1125� 10.
– Compute P (X ≥ 1400) using z-scores or normalcdf:
µ = np = 1375, σ =√np(1− p) ≈ 24.875
P (X ≥ 1400) ≈ normalcdf(1400, 10000, 1375, 24.875) ≈ 0.157
1see Gallup.comMarch 12, 2015 26 / 28
![Page 84: Chapter 5: Sampling DistributionsChapter 5: Sampling Distributions Tuesday (3/10/15): 5.1 {Sampling Distributions {Sample means and the Law of Large Numbers {The Central Limit Theorem](https://reader030.fdocuments.us/reader030/viewer/2022040223/5e5413034285d826427cbc6a/html5/thumbnails/84.jpg)
Approximating Binomial Distributions with Normal
If the number of trials is large compared to either the probability ofsuccess or failure (say both np and n(1− p) are bigger than 10), butstill small compared to the population size, we have
Binom(n, p) ≈ N(np,√np(1− p))
Example: A survey asks a nationwide random sample of 2500 adultstheir opinion on same-sex marriage. Suppose that 55% of Americans 1
actually support same-sex marriage. Estimate the probability that1400 people or more out of the 2500 support same-sex marriage.– Check for validity of normal approximation:np = 2500 · 0.55 = 1375� 10, n(1− p) = 1125� 10.– Compute P (X ≥ 1400) using z-scores or normalcdf:
µ = np = 1375, σ =√np(1− p) ≈ 24.875
P (X ≥ 1400) ≈ normalcdf(1400, 10000, 1375, 24.875) ≈ 0.157
1see Gallup.comMarch 12, 2015 26 / 28
![Page 85: Chapter 5: Sampling DistributionsChapter 5: Sampling Distributions Tuesday (3/10/15): 5.1 {Sampling Distributions {Sample means and the Law of Large Numbers {The Central Limit Theorem](https://reader030.fdocuments.us/reader030/viewer/2022040223/5e5413034285d826427cbc6a/html5/thumbnails/85.jpg)
Sample Proportions
Sometimes, we care more about the proportion of a population thatsatisfies some condition rather than the number of people. For example‘what percentage of people drink coffee every day’?
p̂ =number of successes in sample
size of sample=X
n
– 0 ≤ p̂ ≤ 1, whereas 0 ≤ X ≤ n.
– If X is binomial (or is approximately binomial), then
µp̂ = p
σp̂ =
√p(1− p)
n
– Careful to distinguish p̂, which is a sample proportion, from p, whichis a probability. p̂ is like an approximation of p.
March 12, 2015 27 / 28
![Page 86: Chapter 5: Sampling DistributionsChapter 5: Sampling Distributions Tuesday (3/10/15): 5.1 {Sampling Distributions {Sample means and the Law of Large Numbers {The Central Limit Theorem](https://reader030.fdocuments.us/reader030/viewer/2022040223/5e5413034285d826427cbc6a/html5/thumbnails/86.jpg)
Sample Proportions
Sometimes, we care more about the proportion of a population thatsatisfies some condition rather than the number of people. For example‘what percentage of people drink coffee every day’?
p̂ =number of successes in sample
size of sample=X
n
– 0 ≤ p̂ ≤ 1, whereas 0 ≤ X ≤ n.
– If X is binomial (or is approximately binomial), then
µp̂ = p
σp̂ =
√p(1− p)
n
– Careful to distinguish p̂, which is a sample proportion, from p, whichis a probability. p̂ is like an approximation of p.
March 12, 2015 27 / 28
![Page 87: Chapter 5: Sampling DistributionsChapter 5: Sampling Distributions Tuesday (3/10/15): 5.1 {Sampling Distributions {Sample means and the Law of Large Numbers {The Central Limit Theorem](https://reader030.fdocuments.us/reader030/viewer/2022040223/5e5413034285d826427cbc6a/html5/thumbnails/87.jpg)
Sample Proportions
Sometimes, we care more about the proportion of a population thatsatisfies some condition rather than the number of people. For example‘what percentage of people drink coffee every day’?
p̂ =number of successes in sample
size of sample=X
n
– 0 ≤ p̂ ≤ 1, whereas 0 ≤ X ≤ n.
– If X is binomial (or is approximately binomial), then
µp̂ = p
σp̂ =
√p(1− p)
n
– Careful to distinguish p̂, which is a sample proportion, from p, whichis a probability. p̂ is like an approximation of p.
March 12, 2015 27 / 28
![Page 88: Chapter 5: Sampling DistributionsChapter 5: Sampling Distributions Tuesday (3/10/15): 5.1 {Sampling Distributions {Sample means and the Law of Large Numbers {The Central Limit Theorem](https://reader030.fdocuments.us/reader030/viewer/2022040223/5e5413034285d826427cbc6a/html5/thumbnails/88.jpg)
Sample Proportions
Sometimes, we care more about the proportion of a population thatsatisfies some condition rather than the number of people. For example‘what percentage of people drink coffee every day’?
p̂ =number of successes in sample
size of sample=X
n
– 0 ≤ p̂ ≤ 1, whereas 0 ≤ X ≤ n.
– If X is binomial (or is approximately binomial), then
µp̂ = p
σp̂ =
√p(1− p)
n
– Careful to distinguish p̂, which is a sample proportion, from p, whichis a probability. p̂ is like an approximation of p.
March 12, 2015 27 / 28
![Page 89: Chapter 5: Sampling DistributionsChapter 5: Sampling Distributions Tuesday (3/10/15): 5.1 {Sampling Distributions {Sample means and the Law of Large Numbers {The Central Limit Theorem](https://reader030.fdocuments.us/reader030/viewer/2022040223/5e5413034285d826427cbc6a/html5/thumbnails/89.jpg)
Normal Approximation for Proportion
Making the binomial assumption again - i.e. for sample sizes muchsmaller than the population size (book uses population = 20*sample asa rule-of-thumb), we can use a normal distribution to approximate aproportion:
p̂ ≈ N
(p,
√p(1− p)
n
)
Example: suppose we toss a fair coin 500 times. Let X be the numberof heads.
a) Find the mean and standard deviation of X
b) Find the mean and standard deviation of p̂, the proportion of headsout of 500 tosses.
c) Find the probability that between 45% and 50% of the coins showedheads.
March 12, 2015 28 / 28
![Page 90: Chapter 5: Sampling DistributionsChapter 5: Sampling Distributions Tuesday (3/10/15): 5.1 {Sampling Distributions {Sample means and the Law of Large Numbers {The Central Limit Theorem](https://reader030.fdocuments.us/reader030/viewer/2022040223/5e5413034285d826427cbc6a/html5/thumbnails/90.jpg)
Normal Approximation for Proportion
Making the binomial assumption again - i.e. for sample sizes muchsmaller than the population size (book uses population = 20*sample asa rule-of-thumb), we can use a normal distribution to approximate aproportion:
p̂ ≈ N
(p,
√p(1− p)
n
)
Example: suppose we toss a fair coin 500 times. Let X be the numberof heads.
a) Find the mean and standard deviation of X
b) Find the mean and standard deviation of p̂, the proportion of headsout of 500 tosses.
c) Find the probability that between 45% and 50% of the coins showedheads.
March 12, 2015 28 / 28