Both of them are frequently used to analyze the difference between pretest and post-test results or the increase from the former to the latter. What is the difference?
Deborah J Hilton it is the post-test mean subtracted by the pretest mean divided by the difference between the maximum score subtracted by the pretest mean
Hello Fajar,
Using your explanation of gain analysis, here are the differences:
1. Only the paired/dependent/correlated t-test is directly linked to a specific hypothesis test (e.g., Ho: mu_post - mu_pre = 0, or whatever constant is desired), under which a probability that data as discrepant or more discrepant from Ho would be observed if Ho were true for the population may be determined. As well, it yields a point estimate and estimated standard error of the mean difference (change / gain / / growth / whatever) from observation #1 to observation #2.
2. The gain analysis yields a point estimate of what degree of possible gain (using maximum score as the anchor) was observed for a case or a batch of cases. There is no specific probability distribution associated with these values (unless you are willing make assumptions). As a consequence, there is no direct statistical test available, though some could be applied (such as the sign test, to determine whether negative change is just as likely as positive change, for example).
Is that what you were after?
Good luck with your work.
The term gain score does not specify how the gain is calculated (e.g., post - pre, post/pre, some scaled values) or how the gain scores are to be analyzed. But often it is post - pre and the t-test is done on these, so in that case it would be a paired t-test.
- Badges
- Science method