Suppose I have $N=10$ participants.

For each participant, I have $M=100$ consecutive answers, which are either correct or incorrect. My hypothesis is that the probability of giving a correct answers decreases with time. Can I analyze the data with logistic regression? I would like to do the analysis separately for each participant.


Sure you can. The model then looks like

$$logit(P(trueAnswer)) = \beta_0 + \beta_1 * time$$

with $logit(x)=log(\frac{x}{1-x})$

I would still do a two-sided test for $\beta_1$(since you never know performance may improve over time). If it is negative and significant you have proven your point. If you have not recorded the exact times I think it is still ok to use time = 1,..., M in this case.

It is fine to do the analysis separately per participant since M is large enough, though you may gain power by using a glmm with a random effect for participant.

| cite | improve this answer | |

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.