Slide2 bme adat_2016

Post on 13-Apr-2017

529 views 3 download

Transcript of Slide2 bme adat_2016

session#2 Divényi János @divenyi.janos

b.socrative.com

Student login into room BMEADAT

How to find answersto relevant questions

using data

QUESTION

DATA

ANALYSIS

PRESENTATION

QUESTION

QUESTION

DATA

GET DATA

https://www.quora.com/Where-can-I-find-large-datasets-open-to-the-publichttps://sqlbelle.com/2015/01/16/data-sets-for-bianalyticsvisualization-projects

https://www.quora.com/Where-can-I-find-large-datasets-open-to-the-publichttps://sqlbelle.com/2015/01/16/data-sets-for-bianalyticsvisualization-projects

MEASURE WHAT YOU WANT

http://boredbug.com/wp-content/uploads/2015/08/onesecondbeforedisaster.jpg

http://www.webpages.uidaho.edu/ed571/571-Modules/M3/Sampling_Design-Funny.gif

Prezi & Conversion Rate

Street Bump, Boston

How doesa new piece of

informationaffects

what we knowabout the world?

ww

w.b

igst

ock

phot

o.co

m

Down syndrome screening

Unconditional risk 1:400 = 0.0025Discovery rate 0.83False positive rate 0.047

Unconditional risk 1:400 = 0.0025Discovery rate 0.83False positive rate 0.047

Down if positive?

Down if positive?

Down if positive?

Down if positive?P(+, Down) / (P(+, Down) + P(+, not Down))

Unconditional risk 1:400 = 0.0025Discovery rate 0.83False positive rate 0.047

P(+, Down) / (P(+, Down) + P(+, not Down))

Unconditional risk 1:400 = 0.0025Discovery rate 0.83False positive rate 0.047

0.83*0.0025 / (0.83*0.0025 + 0.047*0.9975)

P(+, Down) / (P(+, Down) + P(+, not Down))

Unconditional risk 1:400 = 0.0025Discovery rate 0.83False positive rate 0.047Down if positive 0.0424

0.83*0.0025 / (0.83*0.0025 + 0.047*0.9975)

P(+, Down) / (P(+, Down) + P(+, not Down))

Unconditional risk 1:400 = 0.0025Discovery rate 0.83False positive rate 0.047Down if positive 0.0424

0.83*0.0025 / (0.83*0.0025 + 0.047*0.9975)

P(+, Down) / (P(+, Down) + P(+, not Down))

Unconditional risk 1:400 = 0.0025Discovery rate 0.83False positive rate 0.047Down if positive 0.0424

0.83*0.0025 / (0.83*0.0025 + 0.047*0.9975)

P(+, Down) / (P(+, Down) + P(+, not Down))

Bayes’ Theorem

P(B|A) = P(A|B) · P(B)

P(A)

P(B|A) =P(A|B) · P(B)

P(A|B) · P(B) + P(A|not B) · P(not B)

1% prevalence

99% accurate test

positive result means risk of

Simplistic example

1% prevalence

99% accurate test

positive result means risk of 50%

Simplistic example

In 1999 Sally Clark was accused for murdering her2 children after she sequentially claimed thatthey died in sudden infant death syndrome (SIDS).

The probability of SIDS is 1 in 8500.

In 1999 Sally Clark was accused for murdering her2 children after she sequentially claimed thatthey died in sudden infant death syndrome (SIDS).

The probability of SIDS is 1 in 8500.

1. If you were the judgewhat other probabilityyou would want to know?

2. Would you convict her?

In 1999 Sally Clark was accused for murdering her2 children after she sequentially claimed thatthey died in sudden infant death syndrome (SIDS).

The probability of SIDS is 1 in 8500.

1. If you were the judgewhat other probabilityyou would want to know?

2. Would you convict her?

3. Do you think she was convicted?

b.socrative.com

Student login into room BMEADAT