hierarchically entering predictors in negbin regression

Hi all,

I am analysing data with no. of lifetime self-injury episodes as the DV. I want to conduct a negative binomial regression and enter my predictors hierarchically with the basic aim of finding the unique contribution of the last predictor in the solution. So basically, my question is whether someone knows how to enter predictors hierarchically when conducting a negative binomial regression.

Thanks a lot in advance!!
Pooja

[Bruce Weaver]

I assume you're using GENLIN.  If you are asking how to make GENLIN do the equivalent of having multiple /ENTER sub-commands when using the REGRESSION command, there is no way to do that.  You'll need to issue a separate GENLIN command for each step in the hierarchy, and you'll have to "roll your own" chi-square test to compare the deviances.  

One other thing:  You'll need to take care to ensure that the same cases are used for all models.  When you run a separate model for each step, there is the possibility of losing cases when you add more explanatory variables (due to missing data on the newly added variables).  I don't have SPSS on this computer, so cannot test this right now, but I believe you can ensure the same cases are used for each model by including ALL of the variables from the fullest model on the first line.  I.e.,

GENLIN Y BY {all categorical variables from fullest model} WITH {all scaled variables from fullest model}
  etc

On the /MODEL sub-command, you can specify which sub-set of those variables to include in the current model.

HTH.

[e-mail] wrote

Hi all,

I am analysing data with no. of lifetime self-injury episodes as the DV. I want to conduct a negative binomial regression and enter my predictors hierarchically with the basic aim of finding the unique contribution of the last predictor in the solution. So basically, my question is whether someone knows how to enter predictors hierarchically when conducting a negative binomial regression.

Thanks a lot in advance!!
Pooja

Thank you!