Chapter 2

You might also like

Download as ppt, pdf, or txt
Download as ppt, pdf, or txt
You are on page 1of 12

DESIGN OF

EXPERIMENTS
Chapter 2
Observational Studies
Observational Studies

• Observational study:
– the subjects assign themselves to control
and treatment groups
• Controlled experiment:
– the investigator assigns the subjects
• Ex. Studies on smoking
– Nobody will smoke for ten years to

please a statistician !
Use of the word “Control”

• A control is a subject who did not get


the treatment
• A controlled experiment is a study
where the investigators decide who
will be in the treatment group and
who will not.
• Controlling for confounding variables.
– HW: Read Chp. 2; smoking example.
The Clofibrate Trial
• A (perfect looking) randomized,
controlled double-blind experiment for
the drug clofibrate to reduce cholesterol
– Result: the treatment is not effective
– Objection: many subject did not take their
medicine anyway !
• Adherers: Subjects who take more than
80% of their medicine.

Comparing Adherers

• Comparing adherers is an
observational study !
Deaths in 5 years
Clofibrate Placebo
Number Deaths Number Deaths

Adherers 708 15% 1,813 15%

Non- adh. 357 25% 882 28%

Total 1,103 20% 2,789 21%


Conclusions of clofibrate
experiment
• Remark: Since the experiment is double-
blind, the psychological bias for adherence is
the same in both groups.
• Conclusions:
i) In both groups only 15% of adherers die
 Clofibrate does not have an effect.
ii) Adherers are different from
non-adherers.
Guesses for how different?
More Examples
• Pellagra: A series of observational studies
and experiments by Goldberger
– Niacin : Causation , Flies : Association
• Cervical Cancer: Circumcision or HPV?
• Ultrasound and low birthweight?
– Observational study: Yes. Confounding
variables adjusted. Still association, why?
– Randomized, controlled experiment: No.
• Samaritans and suicide:
– Observational studies are not experiments!
Controlling for the
confounding factor
• Observational study on sex bias:
Men (44% admitted) Women (30% admit.)
Major
Number Percent Number Percent
A 825 62% 108 82%
B 560 63% 25 68%
C 325 37% 593 34%
D 417 33% 375 35%
E 191 28% 393 24%
F 373 6% 341 7%
Major: confounding variable
• The first two majors were easy to
get into: over 50% of the men applied
• The other four majors were much
harder to get into: over 90% of the
women applied to these majors.
• Table 2 controls for major!
• Simpson’s paradox: Relationships
beween percentages in subgroups
can be reversed, when the subgroups
are combined.
Weighted average
• We can find an admission rate weighted with
respect to admission rate to a major
– Accounts for the confounding variable!
MEN WOMEN IN GENERAL
Number Percent Number Percent Number Men % Women %
825 62% 108 82% 933 62% 82%
560 63% 25 68% 585 63% 68%
325 37% 593 34% 918 37% 34%
417 33% 375 35% 792 33% 35%
191 28% 393 24% 584 28% 24%
373 6% 341 7% 714 6% 7%
Total 2691 1835 4526

Weigh. Avg. 0.445 0.30 0.39 0.43

0.62  933  0.63  585  0.37  918  0.33  792  0.28  584  0.06  714
 0.39
4526
similarly
Confounding
• A difference between the treatment
and control groups –other than the
treatment- which affects the
responses being studied. A
confounder is a third variable,
associated with exposure (treatment)
and with disease.
• Ex. Maybe a gene that causes one to
• smoke and also causes cancer?
Smokers’ dream

You might also like