A statistician responded:
"Since the syntax is identical except for the dependent variable, the issue must be data specific. If one or more of the random effects parameters is noted to be redundant, that could be the source of the problem. This would likely indicate that simplifying the model by reducing the number of random or repeated parameters is indicated. In some cases increasing the number of step halvings or number of Fisher scoring steps can help achieve convergence without Hessian problems, but this isn't guaranteed."
Our advice is to consult with a statistician near you on model and data considerations for this kind of modeling.
------------------------------
Rick Marcantonio
Quality Assurance
IBM
------------------------------
Original Message:
Sent: Tue March 14, 2023 03:08 PM
From: Renee Damstra
Subject: Multi level regression linear Hessian error
The first of the following analysis results in the error in the printscreen, the second does not. What am I supposed to do with this information?
*deelvraag 1: verband tussen interactie met huisdier en expliciete eenzaamheid.
MIXED mood_lonely with pet_interact age gender education SessionInstance
/CRITERIA=CIN(95)
/PRINT=SOLUTION TESTCOV
/METHOD=REML
/FIXED=INTERCEPT pet_interact
/RANDOM=INTERCEPT pet_interact | SUBJECT (subjID)
/Repeated=SessionInstance | SUBJECT (subjID) COVTYPE (AR1).
*deelvraag 1: verband tussen interactie met huisdier en impliciete eenzaamheid.
MIXED Impliciete_eenzaamheid with pet_interact age gender education SessionInstance
/CRITERIA=CIN(95)
/PRINT=SOLUTION TESTCOV
/METHOD=REML
/FIXED=INTERCEPT pet_interact
/RANDOM=INTERCEPT pet_interact | SUBJECT (subjID)
/Repeated=SessionInstance | SUBJECT (subjID) COVTYPE (AR1).
------------------------------
Renee Damstra
------------------------------