10.1 Introduction
- Key Idea 1: Randomization is an important part of collecting high quality data in experiments involving two or more treatments and/or controls...
Discussion
Randomization in this context means that subjects are randomly assigned among treatment groups.
- Key Idea 2: Randomization provides validity, believability, and prediction of accuracy of inference...
Discussion
- A valid study uses proper statistical methodology and thus is truly useful for achieving the purpose of the study.
- Believability means that outside observers will be convinced that the data provide valid information.
- Prediction of accuracy of inference: Inferences from data are almost never 100% certain to be correct. Randomization lets us know the degree of uncertainty, or likely error, in our inference process.
- Key Idea 3: There are two settings for data collection - experiment and survey...
- Key Idea 4: An experiment usually permits random assignment of subjects into treatment and control...
Discussion
For example, in a statistical experiment to determine whether a polio vaccine is effective, half the children are randomly selected to receive the polio vaccine (the treatment group), while the other half do not (control group).
The purpose of random assignment is to be sure the two groups are as alike as possible in all important ways that could influence their likelihood of contracting polio, except for the treatment assignment of who receives and who does not receive the vaccine.
Intuition would have us believe that we could do better than random assignment by using the opinion of experts to balance the two groups. But, in fact, this is worse than randomization.
Randomization automatically does the required balancing, even of important variables that the experts may have failed to think about.
Key Idea 5: In a survey, a real world population is being sampled from...
Key Idea 6: Confounding variables can cause bias...