[Limdep Nlogit List] out of sample prediction

scolombo at ugr.es scolombo at ugr.es
Fri Nov 23 22:39:16 EST 2007

Dear all,

I would like to compare the out of sample prediction capacity of random
parameters and latent class models. Does anyone know if there is any
command that allows to calculate the fitting of an existing model to new
data. (I though to spilt the sample in two, estimate a model in the first
half and used the coefficients to test the fitting in the second half, but
I have not found the way to tell Limdep to use the existing coefficients
and not estimating a new model. Constraining the coefficient seems not a
solution given that 1) I could not constrained all the model coefficients
and 2) in a random paramerter model a constrained coefficienthas not
standard deviation).

Any advice is welcome.


More information about the Limdep mailing list