Cross Validated is a question and answer site for people interested in statistics, machine learning, data analysis, data mining, and data visualization. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

Were finalizing an RCT with two intervention groups (n=13, n=11). Both samples are evaluated pre vs post treatment for pain (VAS), and also against each other (group vs group). However, there are four subjects lost to post treatment in one of the groups (n13 --> n=9). Which statistic model/test would be appropriate to use?

share|improve this question
Could you expand what "RCT" means? – Corone Feb 27 '13 at 10:36
@Corone I am assuming "Randomized Controlled Trial" – Glen_b Feb 27 '13 at 10:42
Randomized controlled trial, yes. – Kristoffer Feb 27 '13 at 10:45
We were initially meant to run a paired t-test to evaluate the decrease in pain for each group, and an unpaired test to detect the differences between the groups. But this seems no longer possible, considering the drop-outs. Am I correct? – Kristoffer Feb 27 '13 at 10:50
You can still do a paired test on the ones that were present in both pre and post and still do unpaired with different sample sizes across groups. – Glen_b Feb 27 '13 at 10:57

Equal sample sizes is not an assumption or requirement of the t-test. Therefore, you can perform them.

The t-test is sensitive to unequal sample sizes. Unequal sample sizes tend to reduce power relative to having the total N divided equally between the groups. Furthermore, while the t-test is robust to unequal variances for alpha, this is NOT the case when the sample sizes are unequal. Then unequal variances can influence alpha quite a lot.

While not necessarily avoiding the potential issues above, which would be minor in the case of small deviations from equal sample sizes, in this case you should probably be doing an ANOVA instead of 3 t-tests. There are a number of ways to handle before-after designs, ANCOVA and factorial ANOVA would be the most popular. I'd recommend the latter. You could test your within effects, between effects, and interaction in a single mixed ANOVA.

share|improve this answer

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.