jan mod posted on Saturday, August 10, 2019 - 6:07 am
Is it correct to do the following calculation in Mplus multilevel logistic regression :
My random intercept is 1.192 (latent continuous) and my level 2 linear coefficient is -0.775. To have a predicted probability of the outcome = 1, I do Exp(1.192-0.775) / (1+ [exp(1.192-0.775)]). So I have the probability for the outcome = 1 for respondents with the zero values in the random intercept + the level two coefficient.
No, with a random intercept its variance also plays a role and you have to use numerical integration over the random intercept to get the correct probability. Your formula is the probability at the mean of the random intercept.