Page 1 of 1

2 level model (logistic regression, IGLS)

Posted: Wed Apr 13, 2011 11:10 am
by jelle

I'm trying to run a 2 level model with a binary response (logistic regression, IGLS). I have no problems when the predictors are level 1 variables (categorical or continuous). However, when trying to run models with level 2 predictors (all continuous data) I encounter the following problems.

Sometimes the model keeps running, so no B and SE are provided. A proposed solution was to allow the level 2 variance to become negative. This solved the problem in some cases. In other cases I recieved the following error messages:
'V gone -ve definite for block 2 iteration aborted (random pass) (moni)' OR '1.#QO'. In the forum I've read that the model might not work for all combinations of predictors / responses. Is this true? Is there no solution for this?

Furthermore, I wonder if it is OK to allow the variance to become negative as the obtained results differ from the results obtained in SPSS?

Thanks in advance!

Best regards