2
$\begingroup$

I have a study about sprint running. Study consists of 16 participants, each performed 7 sprints. This results in a CMC value in evaluating the validity of a new measurement system with 7 per participant. To make it more complicated, every sprint has a CMC value for the start and the full speed phase. Basically, every sprinter has 14 CMC values, 7 for the starts and 7 for the full speed phase. Now, I would like to evaluate if the start and full speed phase differ from each other on the CMC values. I've used a paired samples t-test. However, I'm wondering if I should take the average of the CMC values for each sprint, this would make df=15, or compare all CMC values of every sprint df=111. I know there's probably a better statistical test to use, but I'm quite far into this already and just need to know which option would be the best.

$\endgroup$
1
  • $\begingroup$ Would would be CMC stands for? $\endgroup$ Oct 9 at 12:13
3
$\begingroup$

If you don't average but take the individual values, you will have a problem with dependence, because results for the same participant will be dependent, and this is not taken into account by the test. Not good. This problem goes indeed away by averaging results for each person, but this loses a lot of detailed information (I'd still prefer it if there were no other options). It would be appropriate to set up a model with a random effect for each participant, but you'd need to learn a bit about mixed effects models (of which this would be pretty much the simplest case).

$\endgroup$
1
  • $\begingroup$ Multilevel modelling sounds good. $\endgroup$ Oct 9 at 12:07

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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