1 Approximate Inference 2: Importance Sampling. (Unnormalized) Importance Sampling.
Sampling ion 1
-
Upload
nauman-ur-rasheed -
Category
Documents
-
view
219 -
download
0
Transcript of Sampling ion 1
-
8/6/2019 Sampling ion 1
1/67
1
Chapter 3
Sampling and Sampling Distribution
-
8/6/2019 Sampling ion 1
2/67
2
Introduction:
Collecting data , relying on the entire
population is neither feasible nor practical. Researcher has to select a sample instead
of going in for a complete census .
Inferential or inductive statistics is
primarily concerned with makingconclusions about a certain population
or populations.
-
8/6/2019 Sampling ion 1
3/67
3
On the basis of information obtained from thesample (through sample statistic) , an
inference about the population (populationparameter) is made.
In this process , we need to keep in mindthat the sample contains only a portion of the
population and not the entire population . Soa proper sampling method should be usedfor selecting a sample.
In order to make a good estimate of thepopulation characteristics , selecting areasonably good sampling method is of paramount importance.
-
8/6/2019 Sampling ion 1
4/67
4
Two methods
Census method ( Complete enumeration )
Sample method (Partial enumeration)
-
8/6/2019 Sampling ion 1
5/67
5
Census
100% inspection of the population
Enumeration of each and every unit of population.
It seems to provide more accurate and exact information .For instance , census conducted by Govt of India every 10 years
(Regarding age, martial status, occupation, religion, education,
employment , income, property etc.)
-
8/6/2019 Sampling ion 1
6/67
6
Population
This is a collection of all the units of a specified type defined over agiven space or time
It is defined by :
Content this refers to who or what exactly are the subjects of
interest. Eg. All persons above aged 18 and over
Units this refers to how the subjects are grouped. Eg. Withinhouseholds
Extent this refers to the spatial feature of the population. Eg.
The subjects can only be living in Delhi.
Time this refers to what period of time that your subjects must
possess the particulars named above. Eg. June October 1998
-
8/6/2019 Sampling ion 1
7/67
7
Advantages Accurate
Reliable
-
8/6/2019 Sampling ion 1
8/67
8
Disadvantages
More resources in terms of manpower , money,time and administrative staff etc.
If the test is destructive i.e. the item is destroyedwhile collecting the information about the item ,this option is totally ruled out.
Census method generally time consuming, by the
time results are available it is not of much usedue to changed conditions.
-
8/6/2019 Sampling ion 1
9/67
9
When to use :
Information required about each unit of the
population.
In any manufacturing process in industry, 100%
enumeration should be considered under
following conditions:
1. Serious casualty or loss of life bcoz of defect.
2. A defect may cause loss or serious casulaty.
3. Lot size is small .
-
8/6/2019 Sampling ion 1
10/67
10
Sampling method
Sampling is the most widely used tool for gathering important and useful information fromthe population.
A researcher generally takes a small portion ofthe population foe study, which is referred to assample.
The process of selecting a sample from thepopulation is called sampling.
Sampling over census are defined just in four
word speed, economy, adaptability and scientificapproach.
A properly designed and carefully executedsampling plan yields fairly good results than those
obtained by the census method.
-
8/6/2019 Sampling ion 1
11/67
11
Why is sampling essential?
Sampling saves time.
Sampling saves money
When the research process is destructive in
nature , sampling minimizes the destruction.
Sampling broadens the scope of the study in lightof the scarcity of resources.
It has been noticed that sampling provides more
accurate results , as compared to censusbecause in sampling , non sampling error can becontrolled more easily .
In most cases complete census is not possiblesampling is the only option left.
-
8/6/2019 Sampling ion 1
12/67
12
Merits of sample method over the census:
1. Speed :less time
2. Economy : reduced cost of the enquiry
3. Administrative convenience: less personal staffand limited field of enquiry
4. Reliability :
Sample method contains sampling and nonsampling errors both.
Carefully designed and scientifically executedsample survey gives results more reliable thanthose obtained from a complete census.
-
8/6/2019 Sampling ion 1
13/67
13
Sampling errorsOnly a small portion of the population is studied ,
results are different from census and have certain
amount of error. If the sample is random and
highly representative of the population then alsoerror would always be there. Sampling error is
always due to fluctuations of sampling .
-
8/6/2019 Sampling ion 1
14/67
14
Reasons are:
1. Faulty selection of sample( purposive or judgement
sampling2. Substitution : due to difficulty unit is replaced by anotherof target population
3. Faulty demarcation of sampling units: depends ondiscretion of investigators.
4. Errors due to bias in estimation method
5. Variability of population
-
8/6/2019 Sampling ion 1
15/67
15
In a sample survey these errors can
be controlled by Employing highly qualified ,skilled and
trained personal
Imparting adequate training to theinvestigators for conducting the enquiry.
Better supervision
Using more sophisticated equipment andstatistical techniques for the processing
and analysis of the relatively limited data.
-
8/6/2019 Sampling ion 1
16/67
16
Non-sampling errors
Non-sampling errors are not attributed dueto chance and are a consequence offactors which are within human control. It
presents both in census and sample .
-
8/6/2019 Sampling ion 1
17/67
17
Non-sampling errors
Important factors for non-sampling errors are
1. measuring and recording observations, 2.inaccuracy or incomplete information
3.non response , incomplete response4. training of investigators ,
5. interpretation of questions
6. Lack of coverage
7. Defective method of interviewing and askingquestions.
8. Publication errors
-
8/6/2019 Sampling ion 1
18/67
18
Sampling
Sampling is the process
of selecting a small number of elements
from a larger defined target groupof elements such that
the information gathered
from the small group will allow judgments
to be made about the larger groups
-
8/6/2019 Sampling ion 1
19/67
19
Objectives To obtain the optimum results .
To obtain the best possible estimates of the
population parameters.
-
8/6/2019 Sampling ion 1
20/67
20
Sampling Frame This is a list of the all the units with their
identifications in the target population fromwhich the sample is to be chosen
A subset of subjects for a survey should only betaken from a sampling frame
Generally we identify each unit of the populationby giving it a distinct number, generally from
1,2,3,,N where N is the population size.
-
8/6/2019 Sampling ion 1
21/67
21
Unit of Analysis Sometimes referred to as Sampling Units
A Unit is an element or group of elements ,
living , non- living, on which observationscan be made.
This is the items/units being investigated
A person living in city /household
/employee/a branch/in a bank etc.
Eg : Individuals, households, hospitals
-
8/6/2019 Sampling ion 1
22/67
22
Sampling Units This refers to the items/units selected for
inclusion in the sample
Eg : If Mr. Brown was selected to be
included in the sample then he is a
sampling unit
-
8/6/2019 Sampling ion 1
23/67
23
6L]HRIDVDPSOH
Population size:Normally proportionate
Heterogeneity : More heterogeneity in data , more
the size of sample is required.
Accuracy and Reliability :Bigger size sample
would be more accurate and reliable.
Allocation of Resources : Sample size depends
on the resources allocated . More the resources
(manpower, money, time )are made available,
more the sample size can be increased.
-
8/6/2019 Sampling ion 1
24/67
24
Basics of SamplingT
heoryPopulation
Element
Defined target
population
Sampling frame
Sampling unit
-
8/6/2019 Sampling ion 1
25/67
25
Sampling Methods
Probability
sampling
Non probability
sampling
-
8/6/2019 Sampling ion 1
26/67
26
T
ypes of Sampling MethodsProbability
Simple random
sampling
Systematic random
sampling
Stratified random
sampling Cluster sampling
Non-probability
Convenience
sampling Judgment sampling
Quota sampling
-
8/6/2019 Sampling ion 1
27/67
27
Probability Sampling Probability Samples
Random Samples (though sample units are not chosen
haphazardly)
The probabilities for selecting different samples are
specified
For each unit of the population the probability of it
appearing in any sample is known (I/N: of being selected
in the group ;with replacement)
-
8/6/2019 Sampling ion 1
28/67
28
Example 1 Let a, b and c be three units in the
population , and we want to select a
sample of2
units from the 3 units.
-
8/6/2019 Sampling ion 1
29/67
29
Solution:
Sampling with replacement
Total number of possible samples: Pr
= 3!/1! = 6
Sampling without replacement
Total number of possible samples: Cr
= 3!/(2!)(1!) = 3
-
8/6/2019 Sampling ion 1
30/67
30
There are 3 main steps involved in
choosing a probability sample :
1. Decide on the population of interest
2. Establish a sampling frame
3. Select units from the frame using aprobabilistic algorithm
-
8/6/2019 Sampling ion 1
31/67
31
Simple Random Sampling
Simple random sampling is a method of
probability sampling in whichevery unit has an equal nonzero
chance of being selected
-
8/6/2019 Sampling ion 1
32/67
32
Simple Random Sampling IV
There are two main ways of choosing asimple random sample
1. Table of Random Numbers
2. Lottery Method
-
8/6/2019 Sampling ion 1
33/67
33
Example
Suppose our population consists of
500 units and we have to select a
sample of size5
. In the random number tables, the digits
0 to 9 have equal chance of appearing
in a particular position. The steps of
selecting the sample are as follows:
-
8/6/2019 Sampling ion 1
34/67
34
Procedure:
First identity all sampling unit with 1 to 500
We choose any three columns(row or
diagonal) anywhere in the randomnumber table.
Now we move downwards-selecting 3
digited numbers less than 500 till 5
numbers.
-
8/6/2019 Sampling ion 1
35/67
35
Random Table
12135 65186 86886
49031 45451 07369
70387 53186 97116 93451 53493 56442
74077 66687 45394
-
8/6/2019 Sampling ion 1
36/67
36
Ans
121 490
454
073 453
-
8/6/2019 Sampling ion 1
37/67
37
Systematic (Quasi) Random Sampling
Systematic random sampling is a
method of
probability sampling
in which the defined
target population is ordered
and the sample is selected
according to position using a skip interval
-
8/6/2019 Sampling ion 1
38/67
38
Procedure:
In this method only the 1st unit is
selected at random( by Random
Table). The rest of the units are
selected according to a pattern
depending on a factor which is also
called the sampling ratio.
e.g once in a day, a unit after every 2lots of production
-
8/6/2019 Sampling ion 1
39/67
39
Steps in Drawing a Systematic Random
Sample
1: Obtain a list of units that contains an acceptableframe of the target population
2: Determine the number of units in the list and the
desired sample size3: Compute the skip interval
4: Determine a random start point
5: Beginning at the start point, select the units by
choosing each unit that corresponds to the skipinterval
n
Nk !
-
8/6/2019 Sampling ion 1
40/67
40
Case Let 1
There are 50 employees in an organization
, and each of them has the employee
number from 1 to 50 . We wish to select
10% for assessment of their view on job
satisfaction among the employee of the
organisation . Discuss
-
8/6/2019 Sampling ion 1
41/67
41
Procedure
N = 50; n = 5 = 10% ofN
We may select one random number as
single digit random number varying from 0to 9 say 5.
K = N/ n = 10
Thus the five employee numbers :
5,15,25,35,45 .
-
8/6/2019 Sampling ion 1
42/67
42
Stratified Random Sampling
Stratified random sampling is a
method of probability sampling in which the
population is divided into different subgroups
of non-overlapping homogenous and
Samples are selected from each.
-
8/6/2019 Sampling ion 1
43/67
43
Steps in Drawing a Stratified Random
Sample
1: Divide the target population into
homogeneous subgroups or strata
(strata could be on the basis of geographical
area, different age groups, gender , rural andurban etc. )
2: Draw random samples from each stratum
3: Combine the samples from each stratum into
a single sample of the target population
-
8/6/2019 Sampling ion 1
44/67
44
Stratified Random Sampling III
-
8/6/2019 Sampling ion 1
45/67
45
Eg :
N
N2N1 Nk
n1 n2 nk
-
8/6/2019 Sampling ion 1
46/67
46
Stratified Random Sampling IV
Examples :
Household income or expenditure surveys urban rural
Business surveys employee size
Production
sales
industrial classification
Agricultural surveys
Stratification depends on purpose of survey
-
8/6/2019 Sampling ion 1
47/67
47
Cluster Sampling
Cluster sampling is a method of probability
sampling where the sampling units
are selected on SRS basis in non-overlapping groups
rather than individually.In cluster sampling , the sampling unit is cluster.
But Clusters should be as small as possible consistent
with resources
The number of sampling unit in each cluster should beapproximately same. In stratified sampling , strata
happens to be homogenous but in cluster sampling,
clusters are internally heterogeneous.
-
8/6/2019 Sampling ion 1
48/67
48
Thus cluster sampling involves formation of
suitable clusters of units, and then
selecting a sample of clusters treating them
as unit
-
8/6/2019 Sampling ion 1
49/67
49
Steps
first select a random sample of clusters
from these selected clusters random units
are then selected for study.
-
8/6/2019 Sampling ion 1
50/67
50
Example:
A sample of districts is first selected and
then households are again randomly
chosen from the selected districts.
Cluster sampling is generally used when
the population has natural groupings,
usually in terms of geographical areas.
-
8/6/2019 Sampling ion 1
51/67
51
Case let 2
A survey is conducted among 240 students
of PGDM. They are grouped into 4
divisions of60 each. Then a division could
be considered a cluster. If a sample size is
decided to be 60, then one of the four
cluster could be selected as a sample, and
each of the student of this cluster could beincluded in the sample if n =60.
-
8/6/2019 Sampling ion 1
52/67
52
Non-Probability Sampling
In probability sampling , each unit in the
population or specified group has a chance of
being selected in the sample. The prerequisite in
such sampling is defining the sampling frame i.eidentifying and numbering each and every unit of
the population. However there are certain
situations when it is not feasible and selection is
done on per the convenience . Such samplingcompromising accuracy for convenience , is
referred as Non-probability sampling .
-
8/6/2019 Sampling ion 1
53/67
53
Non-Probability Sampling II
Non-Probability Sampling This involves the selection of a units by arbitrary
methods
The probability of selection for each unit is unknown
It is dangerous to make inferences about the targetpopulation
It is often used to test aspects of a survey such asquestionnaire design rather than make inferencesabout the target population
-
8/6/2019 Sampling ion 1
54/67
54
Convenience sampling
This sampling is also by taking into consideration of
convenience of the investigator .
This method is not very efficient.
It is also used quite a lot in pilot surveys before
,say, launching of a product in the market.
Used for pre-testing of questionnaire.
Convenience sampling relies upon convenience and access
-
8/6/2019 Sampling ion 1
55/67
55
Such sampling is dictated by the needs of
convenience rather than any other
consideration. For opinion poll when one may find it easier
to get the opinion of those in shops, or
restaurants , on pavement rather than
going house to house.
-
8/6/2019 Sampling ion 1
56/67
56
Judgement sampling
In this type of sampling, the investigator decideswhich units to include or exclude in the sample.
Judgement sampling is very simple andconvenient.
It is not as economical as convenience sampling.
Also used for pre-testing of questionnaire. It is useful if the sample size is small.
Judgment sampling relies upon belief
that participants fit characteristics
-
8/6/2019 Sampling ion 1
57/67
57
There is no well-defined scientific method which can
tell us that how one persons judgment is better
than another persons judgment.
Generally , judgment sampling is useful when asample size is small. In case of large samples,
the bias from researchers end may be high.
Disadvantages :
There is scope for personal bias .It may be
influenced by personal bias of the investigator.
-
8/6/2019 Sampling ion 1
58/67
58
Quota sampling
Quota sampling emphasizes representation ofspecific characteristics.,
For example to study eating habits of schoolchildren and college students ,students under18 years.
The quota is fixed due to constraints onavailability of time, cost.
Within the quota stipulated , one has to selecta sample which is representative of thepopulation.
-
8/6/2019 Sampling ion 1
59/67
59
Example
Within the overall quota of interviewing 100
persons for some opinion poll, one may contact
some persons from various categories like college
students, housewives, shopkeepers, office goers ,daily wage earners etc.
In an organization , one might include persons
from all categories of staff cadre-wise as well as
function-wise, department-wise etc.
-
8/6/2019 Sampling ion 1
60/67
60
Advantages
:
The benefits of stratification are available.
Disadvantages :
There is scope for personal bias.
Suitability
It is suitable in marketing research studies
where it is not possible to stick to it without
delay and expenditure.
-
8/6/2019 Sampling ion 1
61/67
61
Sampling and Non Sampling
Errors
Sampling errors arise from the fact that a
sample has been used to study the population.
These errors are generally not present in a
complete census as they are associated withthe process of selecting a sample.
A sampling error is the difference between the
estimate obtained from sampling and the truevalue for the entire population.
-
8/6/2019 Sampling ion 1
62/67
62
Sampling errors may be due to
faulty selection of the sample,
improper data collection method,
Human beings limitation of recording data
due to lack of competence, training or human
fatigue.
incorrect methodology of analyzing the data
Wrong calculation
-
8/6/2019 Sampling ion 1
63/67
63
Sampling Errors
Sample size
-
8/6/2019 Sampling ion 1
64/67
64
Non-sampling errors are errors arising
during the course of all survey
activities other than sampling.
They exist both in sample surveys as well
as censuses.
-
8/6/2019 Sampling ion 1
65/67
65
Figure 1
Sample Size
Non-Sampling
Error
-
8/6/2019 Sampling ion 1
66/67
66
Factors to Consider in Sample Design
Research objectives Degree of accuracy
Resources Time frame
Knowledge of
target population Research scope
Statistical analysis needs
Steps in Developing a Sampling
-
8/6/2019 Sampling ion 1
67/67
67
1. Define the Target Population
3. Identify the Sampling Frame(s) Needed
4. Identify the Appropriate Sampling Method
5. Determine Sample Sizes and Contact Rates
6. Create Plan forSelecting Sampling Units
7. Execute the Operational Plan
Steps in Developing a Sampling
Plan
2. Select the Data Collection Method