If I have two groups (those who receive treatment and those who do not) and I test them before and after the treatment, what is the best statistical test to use?
I was looking at unpaired t-test to compare the treatment and non-treatment groups and paired t-test to compare before and after treatment, however wondering if there is another test that would also be suitable?
Hello Julie Pom In my opinion, the best method of statistical analysis in this study is the use of analysis of covariance (ANCOVA) in which the before measurements are entered as covariates and after measurements as dependent variable. The group is also analyzed as a fixed factor.
If you wish to adjust for covariates (age, sex, ethnicity, education, etc.), I would recommend that you use linear mixed effect model and contrast estimated marginal means. Not sure how it can be done with SPSS, but it shouldn't be terribly difficult.
Dear Julie:
For comparing a variable which you desire to assess during a time period, first of all you should determine the kind of variable (categorical/continuous) and the distribution status of your desired variable (normal/abnormal).
If your variable is continuous with normal distribution, you can use paired t test.
Otherwise, you have to use other specific tests like McNemar or Wilcoxon tests based on the characteristics of your desired variable.
Of note, if you use SPSS software, you should first select the cases who underwent treatment and perform your analysis separately and do it again for your control subjects.
Regards
- Badges
- Science topic
- Similar topics
- Linguistics
- Composition Studies
- Publication