Hi all
Thank you all in advance for your valuable comments.
I’ve analyzed my data using Repeated-Measures ANOVA but there is an issue that I can’t get.. In my study, I had two different groups (exercise and control) and also exercise intervention (pre and post). Shortly, in the study, I tried to reveal the effects of an exercise model on fat oxidation during exercise. To analyze my data, I’ve used Repeated-Measures ANOVA. But once analyzing the data, Repeated-Measures ANOVA shows that there is a significant difference in time (p=.025) but no differences in between the groups (time*gruops). But when I want to see which group has significant difference compared to baseline using Paired Samples T-test, there is no difference in either group. So how should I write my results or should I analyze my date with another statistical method?
Thank you
Pre-post is a repeated measure test that ignores everything else that repeated ANOVA takes into account. Best wishes, David Booth
"non-significant" is not implying "no difference".
This is the "absence of evidence is taken as evidence of absence" fallacy.
A possibility could be to compute the difference between pre and post in both samples and use a Student's t-test. It would also give you a descriptive idea of the data.
Best
John-Kåre
Thank you,
I really appreciate your comments. Just want to ask again not to make mistake. Can I submit my results by expressing that we found significantly difference after analysing data with Repeated-Measures AVONA but paired sample t-test showed that there was no difference between pre and post tests within the groups?
Yes, you can, although I don't think that it is smart to use paired t-tests on only a part of the data. Instead you might consider a repeated-measures one-factorial ANOVA (with 4 levels representing the four different combinations of time and group), in which you make the comparisons of interest. This will ensure that all the data is used to estimate the standard errors and the tests are performed with the same degrees of freedom.
If I get this right, you seem to have a significant difference from pre to post data in the whole sample, but no difference at the group level. It may be due to your sample size and the possibility to detect difference between intervention and control. How large was your sample size? It is also very often that a control group have an improvement without any intervention given.
You should check descriptively and try to find out "what is going on" in your data.
If your study is a RCT you could merely test outcomes (at post-test) between groups with a simple t-test (your randomization should account for differences at baseline - also differences on the main outcome measure).
Best
JK
It is unclear that how many times the measurements were taken to apply RM-ANOVA, and how many times the post level was compared with pre/baseline measurement. Did any correction on p value to account for multi comparison was done?.
Hello. If I understand it correctly, you had a time effect (you said "p=.025"). It implies that all subjects, no matter which group they belong to, changed from baseline. It means if you pool all your subjects in just baseline and post-treatment measurements, the test will state there's significant difference (as reported for your p value). However, you also need to report if there was group effect, i.e. pooled subjects in exercise and control groups, no matter the time. Finally, I don´t see the problem with your analysis. Your ANOVA showed non-significant interaction (time*group) (i.e. comparing both groups at every time point) and your post hoc test showed the same. I think you have a different interpretation rather a problem with your results, because both ANOVA and post hoc test showed the same results (and actually it wouldn't be necessary to perform a post hoc).
I hope it would be useful.
- Badges
- Science topic