Things to Watch for in Various Types of Studies
In addition to all the problems of designing good measurement instruments
(especially when people are involved), each type of study has its own
pitfalls to avoid. Proper statistical design of these studies will
take these into account. A good report of a statistical study should
convince you that this has been done.
All the problems here can be summarized in one problem:
failing to get a representative sample.
There are several ways this can occur.
- Wrong sampling frame
- Not reaching selected individuals
- Low response rate
- Volunteer sample
- Convenience sample
- Confounding variables
- Interacting variables
- Placebo, Hawthorne, and experimenter effects
- Ecological validity and generalization
Confounding variables and difficulties in determining cause and effect,
Using the past or self-reporting as a source of data
Seeing Through Statistics, 2nd ed. by Jessica Utts,
pages 58, 77, and 84.
This page maintained by:
Thursday, 30-Aug-2001 12:17:13 EDT
Department of Mathematics and Statistics