1
$\begingroup$

I have data corresponding to reports of "mind wandering", i.e. thinking about something unrelated to the task, and I want to look at the link between these reports and trust (regarding a system capability). For each subject, I have 25 values of mind wandering (binary variable) and trust (range between 1 and 5).

I firstly used a Kendall rank correlation, however someone told me that it would not account for repeated measures. That same person told me to use logistic models.

I am not entirely sure that looking at logistic models would tell me the same thing as correlation test. However I have never done logistic regression, and so I am a bit lost with the possibilities offered by the technic. Is there a possibility to make correlations accounting for repeated measures?

$\endgroup$

1 Answer 1

-1
$\begingroup$

Yes, you can separate the subject data samples and compute the means, then calculate the correlations between the means.

$\endgroup$
5
  • $\begingroup$ Thank you for this answer. I agree that this would solve the problem, however I think it would lower the sharpness of our test by aggregating (and therefore loosing) much data. $\endgroup$
    – Pyxel
    Jun 22, 2018 at 14:45
  • $\begingroup$ Have you tried the split-plot ANOVA, what about MANOVA and /or Mixed models approached $\endgroup$ Jun 22, 2018 at 15:46
  • $\begingroup$ Thank you for your comment! However, as I explain in the original post, I don't know if the results would be similar to a correlation. For example, if I have a significant predictor in my mixed effect model, I can say that this predictor influences my dependent variable (and only linearly). However can I say that my variable and my predictor are correlated? I don't think so. $\endgroup$
    – Pyxel
    Jun 23, 2018 at 16:47
  • $\begingroup$ If the predictor can directly and linearly influence the dependent variable , then correlation exists. Only check the magnitude $\endgroup$ Jun 25, 2018 at 5:49
  • $\begingroup$ I agree with you on that, however a correlation can also exist without linear influence (e.g. a quadratic relation would be revealed by a correlation test, but not by a linear model). $\endgroup$
    – Pyxel
    Jun 26, 2018 at 8:28

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.