 Binary Growth Model    Message/Author  Jaime Maerten-Rivera posted on Monday, January 05, 2015 - 11:41 am
I ran a binary growth model using logit regression with 7 waves of data which will later be used in a two-part semicontinuous model. I am trying to calculate the probability of using alcohol and want to make sure that I am doing this correctly.

The relevant results are listed below.

MEANS
Int 0.00
Slp 0.684

Thresholds 3.346

Variances
Int 7.219
Slp 0.330

P(Y=1| x) = 1/[1+exp(-logit)],

where the logit = -threshold +mean(intercept) + t(slope)

Therefore for time=0,
Logit =-(3.346) + (0 + 0*0.684)= -3.346
P(Y=1|x) = 1/[1+exp(-3.346) = .034

And for time=1,
Logit =-(3.346) + (0 + 1*0.684)= -2.662
P(Y=1|x) = 1/[1+exp(-2.662) = .065

Is this correct? Also, you mention in a previous post using the Probability output and graphics in MPlus to graph this? How can I do this?  Bengt O. Muthen posted on Monday, January 05, 2015 - 5:20 pm
This probability involves numerical integration to integrate out the random effects, that is, the intercept and slope latent variables. Because of this, it is not straightforward to computed this yourself. You can compute the probability conditional on the means of i and s if you like - this does not involve integration.    Topics | Tree View | Search | Help/Instructions | Program Credits Administration