Scribd
Upload
28 views15 pages

Quantitative Reasoning: Design of Studies: Smoking and Health

An observational study differs from a controlled experiment in that investigators do not assign subjects to treatment and control groups. An association between two characteristics A and B in a population is shown if either (1) the rate of A given B is not equal to the rate of A given not B, or (2) the rate of B given A is not equal to the rate of B given not A. The Consistency Rule states that if one of these is true, the other must also be true, so it is sufficient to check just one to demonstrate an association.

Uploaded by

jzhong_7
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
28 views15 pages

Quantitative Reasoning: Design of Studies: Smoking and Health

An observational study differs from a controlled experiment in that investigators do not assign subjects to treatment and control groups. An association between two characteristics A and B in a population is shown if either (1) the rate of A given B is not equal to the rate of A given not B, or (2) the rate of B given A is not equal to the rate of B given not A. The Consistency Rule states that if one of these is true, the other must also be true, so it is sufficient to check just one to demonstrate an association.

Uploaded by

jzhong_7
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 15

Quantitative Reasoning

Design of Studies: Smoking and Health

Controlled Experiment vs Observational


Study

Assignment by

Controlled
Experiments
Investigators

Observational
Studies
Subjects

Long-term effects of smoking can only be investigated


through observational studies.

Smoking and heart disease

A city has 100,000 residents.

Among 15,000 smokers, 38 have heart disease.

Among 85,000 non-smokers, 44 have heart disease.


Number
Smokers (15,000)

38

Non-smokers (85,000)

44

Association

Rate of heart disease among . . .

38
Smokers:
100% = 0.25%
15,000
44
Non-smokers:
100% = 0.05%
85,000
Smokers (15,000)
Non-smokers (85,000)

Number
38
44

Rate
0.25%
0.05%

Association between smoking and heart disease

Association btw smoking & heart disease

Salk randomised experiment

Treatment
Control

Rate
(per 100,000)
28
71

NFIP study

Treatment
Control

Rate
(per 100,000)
25
54

Association is not Causation


Study

Association due to

Salk randomised experiment

Vaccine only.

NFIP

Vaccine, family background, other


factors.
Smoking? Other factors?

Smoking and heart disease

Two-by-two table
Smokers
Non-smokers
Column total

Heart disease No heart disease


38
14,962
44
84,956
82
99,918

Rate of heart disease among:

38
Smokers:
100% = 0.25%
15,000
44
Non-smokers:
100% = 0.05%
85,000

Row total
15,000
85,000
100,000

Flipped rates
Smokers
Non-smokers
Column total

Heart disease (HD)


38

No heart disease
14,962

Row total
15,000

44
82

84,956
99,918

85,000
100,000

Rate of smoking among HD:

38
82

Rate of HD among smokers:

38
15,000

Shorthand: rate(smoke | HD) = rate of smoking among HD,

100% = 46.3%
100% = 0.25%

rate(HD | smoke) = rate of HD among smokers.

In general, rate(smoke | HD) rate(HD | smoke). True for any


2x2 table.

Two views on association


HD
38
44
82

Smokers
Non-smokers
Column total

rate(smoke | no HD) =

38
82

rate(smoke | HD) =

14,962
99,918

No HD
14,962
84,956
99,918

Row total
15,000
85,000
100,000

100% = 15.0%.

100% = 46.3%.

Difference confirms association between smoking and HD, just like


comparing rate(HD | smoke) = 0.25% and rate(HD | not smoke) =
0.05%.
Both views are valid, though one may seem more natural.

IQ and wealth
Low IQ (< 90)

Poverty
12

Adequacy
48

Row total
60

High IQ (> 90)


Column total

10
22

170
218

180
240

Adult US population; numbers in millions.

(1) rate(poverty | low) = 20.0%, rate(poverty | high) = 5.6%.


(2) rate(low | poverty) = 54.5%, rate(low | adequacy) = 22.0%
(1) and (2): equally valid evidence of association.

Flipped rates are unequal: rate(poverty | low) vs rate(low | poverty).

A Consistency Rule

Let A and B be characteristics. For example, A = heart


disease, B = smoking. Or A = poverty, B = low IQ.

Two statements:
(1) rate(A | B) rate(A | not B)
(2) rate(B | A) rate(B | not A).
Both true in examples; not a coincidence.

The Consistency Rule: if (1) is true, then (2) must also be true;
and vice versa.

To show an association between A and B, it is enough to


check just (1) or (2).

Is there an association?

Hard to observe no association: rate(A | B) = rate(A | no B).

What if the rates differ very slightly? Whether this matters


depends on the situation. Not an issue in module.

Summary

How is an observational study different from a controlled


experiment?

What does it mean to say that two characteristics A and B are


associated in a population?
In order to show an association between A and B, it is enough
to check just one of the following:
(1) rate(A | B) rate(A | not B), or
(2) rate(B | A) rate(B | not A).
Justified by the Consistency Rule.

You might also like

pFad - Phonifier reborn

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.

Note: This service is not intended for secure transactions such as banking, social media, email, or purchasing. Use at your own risk. We assume no liability whatsoever for broken pages.


Alternative Proxies:

Alternative Proxy

pFad Proxy

pFad v3 Proxy

pFad v4 Proxy