Convergence criterion not satisfied in logistic regression

Aline1610 — Wed, 21 Jul 2021 10:19:54 GMT

Hello,

I am performing an univariate conditional logistic regression for my nested case-control study. My dependent variable is the lymphoma status and my indepedent variable is the energy intake with 3 classes including a "missing values" category.

When i generated the PROC LOGISTIC model with the STRATA function, this warning appeared :

"ERROR: NRRIDG Optimization cannot be completed.
ERROR: Optimization routine cannot improve the function value. You may want to change the convergence criteria (ABSFCONV=,
FCONV=, GCONV=, XCONV= options) in the MODEL statement"

When i removed the "missing values" category the warning disappeared. I dont understand why with this variable only i have this problem. I already tested many other variables with a "missing values" category and the proc logistic models worked well. What should i do ? It asks me to change the convergence criteria but i have no clue what does this mean...

Thanks in advance for your help !

Aline.

Re: Convergence criterion not satisfied in logistic regression

PaigeMiller — Wed, 21 Jul 2021 12:54:42 GMT

It is possible that when the value of energy intake is MISSING then it only appears with one value of lymphoma status, and so this could cause problems. You should check this.

topic Re: Convergence criterion not satisfied in logistic regression in Statistical Procedures

Convergence criterion not satisfied in logistic regression

Re: Convergence criterion not satisfied in logistic regression