I'm currently conducting a study with two objectives. The primary objective is to assess the influence of using different recovery modalities on repeated sprint performance in youth footballers. To do this, I am analysing performance (%decrement, average sprint, and recovery duration) during repeated sprints under three conditions (lets just call them 1,2 and 3). For this alone I'm assuming I would use a repeated measures ANOVA. However, the secondary objective of the study is to assess the influence of stage of maturation on performance during repeated sprints under the same three conditions. To do this, I have split the participants up into a more and less mature group.

Therefore, as well as assessing differences within the group as a whole (i.e. without splitting them up, condition 1 vs 2 vs 3), I am also to assess between group differences (i.e. more vs less mature), as well as within subject difference within the group (i.e. condition 1 vs 2 vs 3 in the more and less mature group).

Would a mixed ANOVA be more appropriate to indicate differences within subject and between groups? Would I have to perform a separate ANOVA to assess within subject differences in the group as a whole?


1 Answer 1


Yes, a mixed model is the way to go. In particular, this is a so-called split-plot layout: Each measurements is influenced by three fixed factors: Maturity --the whole plot factor--, sprint condition --the sub-plot factor-- and the interaction between both. An interaction would indicate that maturity also determines the differences (in sprint time, recovery duration,...) between the conditions. So first you have to test for the interaction. If there is a significant interaction, you will in fact have to do seperate analyses, because an interaction indicates that a general conclusion about only maturity or only condition is not possible.

  • $\begingroup$ Thanks very much for your answer. Will this also give me the within subject differences for performance variables for the group as a whole (i.e. forgetting the maturity groups and simply looking at the influence of the different sprint conditions)? My thinking was that I may have to do a seperate repeated measures ANOVA. Also, in two of the 3 trials, recovery was self-guided, and we are looking at differences in the recovery durations within subject and between group. However, given that there was only 2 measures of recovery, would this simply be analysed using paired and indepdnent t-tests? $\endgroup$
    – user73130
    Apr 9, 2015 at 13:07
  • $\begingroup$ Yes, as usually in statistical software, the results for all ANOVA factors are printed out in one table. So the results for the effect of the conditions alone will also be calculated. But it is not really "forgetting the maturity groups". As this information is present, it will be used to get more precise effect estimations. $\endgroup$ Apr 9, 2015 at 13:24
  • $\begingroup$ Your 2nd question: Usually the subjects are random factors, i.e. something you want to generalize about. You can consider them as fixed factors (e.g. "did Bob run as fast as Jeff?"), but then you need at least two identical replications for testing, say the two conditions, if they are really identical. Still, two replications are very few for a meaningful result. $\endgroup$ Apr 9, 2015 at 13:31

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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