|
|
 |
| Cross lag with count data: neg binomi... |
 
|
| Message/Author |
|
|
|
|
Hello,
I am analyzing data using a cross lag panel model where all variables are zero-inflated count.
I am getting the following error:
*** FATAL ERROR
NEGATIVE BINOMIAL VARIABLE HAS IMPUTED VALUE GREATER THAN 50000.
What can I do to fix this? I get this whether or not i do zero-inflation.
Background:
In each wave, two variables are considered (x and y); in each case, they are predicted by X and Y from the previous wave. I also want to model the residual covariance between X and Y at each timepjoint; prior browsing on this forum suggested I should make a latent variable for each wave that is indicated by X and Y (see below). I have many waves in the model; the key questions are the size and significance of the cross-lag slopes.
x_08 on x_04 y_04;
y_08 on x_04 y_04;
x_12 on x_08 y_08;
y_12 on x_08 y_08;
... etc
f1 by x_08@1 y_08;
f2 by x_12@1 y_12;
... etc.
f1@1; [f1@0];
f2@1; [f2@0];
... etc |
|
|
|
|
First, check your data so that you don't have huge counts.
Second, explore the reason for the problem by running one regression at a time.
Note also that when used as predictors, a negbin variable is treated as continuous. There is no y* version for counts. |
|
|
| Back to top |
|
|
