I understand that if you collect data pre and post-intervention from the same individuals, a paired sample t-test is the best option to compare differences between the scores.
However, if a few of the subjects completed the pre but not the post intervention tests, meaning you now have unequal sample sizes pre and post, and do not wish to remove their data, which statistical test is best to use? e.g. Pre n = 25, post n = 22.
Is a student t-test the correct option?
.
partially paired t-test ? it does exist ; see :
Article Notes on two sample tests for partially correlated (paired) data
Article A Simple and Robust Method for Partially Matched Samples Usi...
(review carefully the hypothesis before use ...)
.
If you include the data where there is no post-treatment response, then you no longer have a paired test. However, the unpaired test (e.g. t-test) will assume that the experimental units in the post-treatment group are uncorrelated with the pre-treatment group and that is clearly not the case. The experiment for t-test would be where you have 50 people, 25 placebo and 25 treatment where 2 people in the treatment had to be removed for some reason.
One option is to run an analysis using the paired data and remove the two subjects that did not finish the test. Then run another analysis only on the pre-treatment subjects.
There is no information with the individuals that did not do both pre and post. Use only the ones that did both in a paired t test. Best, David Booth
There is the population that was sampled (n=25). Of that there is a subpopulation that did not finish the experiment. The question is whether the subpopulation that did not finish the experiment is any different than the population that was sampled.
Of course with only three individuals in the subpopulation that did not finish the experiment it will be difficult to see if they differ from the sampled population.
Of course if the people that did not complete the experiment did so because they moved, died, or some medical condition then simply describe why these individuals left the experiment and follow David's suggestion to delete them.
Dear Fabrice, Timothy and David,
Thank you all very much for your suggestions and information on how I can analyse my data going forward. I really appreciate your advice.
Kind regards,
Kiera
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Sampling Design
- Sample Size