Experimental design

Basic principles

1. Formulate question/goal in advance

2. Comparison/control
3. Replication
4. Randomization
5. Stratification (aka blocking)
6. Factorial experiments

2
 Example

Question: Does salted drinking water affect blood

pressure (BP) in mice?

Experiment:
1. Provide a mouse with water containing 1% NaCl.
2. Wait 14 days.
3. Measure BP.

Comparison/control

Good experiments are comparative.

• Compare BP in mice fed salt water to BP in mice

fed plain water.
• Compare BP in strain A mice fed salt water to BP
in strain B mice fed salt water.

Ideally, the experimental group is compared to

concurrent controls (rather than to historical controls).

4
 Replication

Why replicate?

• Reduce the effect of uncontrolled variation

(i.e., increase precision).
• Quantify uncertainty.

A related point:
An estimate is of no value without some
statement of the uncertainty in the estimate.

6
 Randomization

Experimental subjects (“units”) should be assigned to

treatment groups at random.

At random does not mean haphazardly.

One needs to explicitly randomize using

• A computer, or
• Coins, dice or cards.

Why randomize?

• Avoid bias.
– For example: the first six mice you grab may have
intrinsically higher BP.
• Control the role of chance.
– Randomization allows the later use of probability
theory, and so gives a solid foundation for
statistical analysis.

8
 Stratification

• Suppose that some BP measurements will be made

in the morning and some in the afternoon.
• If you anticipate a difference between morning and
afternoon measurements:
– Ensure that within each period, there are equal
numbers of subjects in each treatment group.
– Take account of the difference between periods in
your analysis.
• This is sometimes called “blocking”.

Example

• 20 male mice and 20 female mice.

• Half to be treated; the other half left untreated.
• Can only work with 4 mice per day.

Question: How to assign individuals to treatment

groups and to days?

10
 An extremely
bad design

11

Randomized

12
 A stratified design

13

Randomization
and stratification
• If you can (and want to), fix a variable.
– e.g., use only 8 week old male mice from a single
strain.
• If you don’t fix a variable, stratify it.
– e.g., use both 8 week and 12 week old male mice,
and stratify with respect to age.
• If you can neither fix nor stratify a variable, randomize it.

14
 Factorial
experiments
Suppose we are interested in the effect of both salt
water and a high-fat diet on blood pressure.
Ideally: look at all 4 treatments in one experiment.
Plain water Normal diet
Salt water High-fat diet

Why?
– We can learn more.
– More efficient than doing all single-factor
experiments.
15

Interactions

16
 Other points

• Blinding
– Measurements made by people can be influenced
by unconscious biases.
– Ideally, dissections and measurements should be
made without knowledge of the treatment applied.
• Internal controls
– It can be useful to use the subjects themselves as
their own controls (e.g., consider the response
after vs. before treatment).
– Why? Increased precision.

17

Other points

• Representativeness
– Are the subjects/tissues you are studying really
representative of the population you want to
study?
– Ideally, your study material is a random sample
from the population of interest.

18
 Summary

Characteristics of good experiments:

• Unbiased • Wide range of applicability
– Randomization – Deliberate variation
– Blinding – Factorial designs
• High precision • Able to estimate uncertainty
– Uniform material
– Replication
– Replication
– Randomization
– Stratification
• Simple
– Protect against mistakes

19

Salk vaccine trial

1916: first polio epidemic in the US

next 40 years: hundreds of thousands of victims

By 1950s: several vaccines developed; that by Jonas Salk

appears most promising

1954: Public Health Service and Nat’l Fdn for Infantile Paralysis
(NFIP) ready to test the Salk vaccine in a field trial

See Freedman, Psiani, Purves (1998)

Statistics, 3rd ed, Ch 1–2

20
 Possible designs for the vaccine trial

1. Give the vaccine to many children and look at the rate vs the
previous year.
2. Compare those vaccinated to those whose parents refused
vaccination.
3. Vaccinate grade 2 (in consenting) and compare to grades 1
and 3. [This is what the NFIP chose to do.]
4. Vaccinate some portion (chosen at random) of those whose
parents consent.

Best study:
double-blind randomized placebo-controlled

21

Results of 1954 Salk vaccine trial

The randomized controlled

double-blind experiment
Size Rate
Treatment 200,000 28
Control 200,000 71
No consent 350,000 46

The NFIP study

Size Rate
Grade 2 (vaccine) 225,000 25
Grades 1 & 3 (control) 725,000 54
Grade 2 (no consent) 125,000 44
Note: Rates are per 100,000

22
 Points

• NFIP study: vaccine appears to lower rate 54 → 25 (vs 71 →

28).
The control group included children whose parents would not
have consented.
• Might the vaccine have no effect? (Could the observed differ-
ences be simply chance variation?)
– In the randomized controlled trial, it is relatively simple to answer this
question, as the role of chance was according to our design.
– In the NFIP study, it is impossible to tell, as chance is not under our
control.

23

The portacaval shunt

A long, hazardous surgery to treat cirrhosis of the liver.

Do the benefits outweigh the risks?
Over 50 studies have considered this.

Degree of enthusiasm
Design Marked Moderate None
No controls 24 7 1
Controls, but not randomized 10 3 2
Randomized controlled 0 1 3

In the studies where the controls were not chosen at random,

sicker patients were chosen as controls.

24
 Historical controls

Historical controls: patients treated the old way in the past.

Problem: treatment group and historical control group may differ in
important ways besides the treatment.

Randomized Historically
controlled controlled
+ – + –
Coronary bypass surgery 1 7 16 5
5-FU 1 7 2 0
BCG 2 2 4 0
DES 0 3 5 0

25