Experiments vs. Observational Studies
Do NOT influence the response
- Has global warming effected penguin mating behavior
Do something to your individuals
- Does housing penguins in warmer environments effect mating behavior?
The phenomenon where patients get better because they expect the treatment to work.
Many statistical studies involve testing the effectiveness of drugs. A placebo looks identical to the actual drug but contains no active ingredient and so has no real physical effect.
Humans want to be helped by the medication that is administered to them. If they think they are receiving a drug to help their condition, they tend to improve even if it turns out that the drug is a placebo.
Characteristics of a Well-Designed and Well-Conducted Experiment
The effect of lurking variables, most often by comparing treatments
Example: a "Control group" in a drug study to eliminate the "confounding effects" of environment or the placebo effect
Each treatment on many units to reduce chance variation
Example: do the mouse study many times
Use probability (chance) to assign experimental units to treatments
May be the most important!!
Because it allows us to say the different treatment groups start out similar
Completely Randomized Design
If all the experimental units (subjects of the experiment) are randomly assigned to either the control group or to the treatment group, then the experiment has a completely randomized design.
Randomize by assigning each subject a number and then generating it to choose treatment groups
Placing subjects into groups of similar individuals. The random assignments into treatment groups is carried out separately within each block (think stratified random sample)
Matched Pairs Design
Subjects are matched into pairs and get different treatments
Matched pairs are more similar than random unmatched subjects
Randomizing the rest of the experiment is still important!!!
Experimental Set Up
Treatment Imposed = Independent Variable = Factors
Experimental Units = Subjects
Response Variable Observed = Dependent Variable
In a double-blind experiment, neither the subjects nor the researchers know to which group, treatment, or control, subjects have been assigned. If a researcher knows that a subject is in the control group, they do not expect a treatment effect, and their measurement of a response might be understated. If a researcher knows that a subject is in the treatment group, they might overstate a response simply because they expect it.
An experiment might also be single-blind. In this case, only one of the participants, either the subjects or the researchers, knows to which group the subjects have been assigned.
Avoids unconscious bias
Generalizability of Results
To determine if our data is "statically significant"
- i.e. is an observed effect so large that it would rarely occur by chance
If we designed and conducted our experiment well, we can generalize these results to the population!
Control groups are used in experiments in order to
Control the effects of outside variables on the outcome
Control the subjects of a study to ensure that all participate equally
Guarantee that someone other than the investigators, who have a vested interest in the outcome, controls how the experiment is conducted
Achieve a proper and uniform level of randomization
Angiotensin-converting enzyme (ACE) inhibitors are used to treat high blood pressure. We want to conduct an experiment to see if a new blood pressure drug works even better than ACE inhibitors. Design a completely randomized experiment to test this.
In conducting an experiment to see if a new blood pressure drug works even better than ACE inhibitors. We learn that men and women may react differently to common cardiovascular drug treatments. Design a randomized experiment to test this with your new information on gender.
- We will conduct a randomized blocked experiment, blocking on gender.
The Community Intervention Trial for Smoking Cessation (COMMIT) asked whether a community-wide advertising campaign would reduce smoking. The researchers located 11 pairs of communities that were similar in location, size, economic status, and so on. One community in each pair participated in the advertising campaign and the other did not. This is
an observational study
a matched pairs experiment
a completely randomized experiment
a block design
A study of cell phones and the risk of brain cancer looked at a group of 469 people who have brain cancer. The investigators matched each cancer patient with a person of the same sex, age, and race who did not have brain cancer, then asked about the use of cell phones. This is
an observational study
an uncontrolled experiment
a randomized comparative experiment
a matched pairs experiment
A fitness instructor wants to test the effectiveness of a performance-enhancing herbal supplement. Design an experiment to test this supplement
Double blind, (placebo controlled), matched pairs experiment:
Match subjects based on performance in a fitness test and gender
Randomize who in the pair gets the new supplement and who gets the old supplement / placebo.
Give both in the same packaging, making sure the subject doesn't know the group and person measuring fitness doesn't know the group either.
A researcher believes that students may do better on a test when taken in the same classroom where the material was learned. To test this theory she plans to present a lecture and then give students a multiple choice quiz on the material. She knows there is a lot of variability in the students' academic ability. Design a study to test her hypothesis.
- Block on academic ability