I am fitting a multinomial logit model with a random intercept in PROC GLIMMIX, and one of my F-statistics in the Type III table is infinite. When I use the NLOPTIONS statement to change the convergence criterion to an integer (like 50), the model converges and I get an F-statistic that more along the lines of what I would expect to see and my results make sense. However, I do not know what the default value of FCONV is when I am fitting the models. Clearly, it must be very small. I cannot find this in the documentation.
The doc is here
SAS Help Center: Syntax: NLOPTIONS Statement
If you haven't changed the default value of the FDIGITS option, then the default value for FCONV is approximately machine epsilon, which is ~2E-16.
If you set FCONV=50, you are not getting convergence. Instead, you are saying that the algorithm should quit iterating even though the relative change of the objective function is about 50. Although you might get parameter estimates, the estimates are not very good since the model does not fit the data well.
The doc is here
SAS Help Center: Syntax: NLOPTIONS Statement
If you haven't changed the default value of the FDIGITS option, then the default value for FCONV is approximately machine epsilon, which is ~2E-16.
If you set FCONV=50, you are not getting convergence. Instead, you are saying that the algorithm should quit iterating even though the relative change of the objective function is about 50. Although you might get parameter estimates, the estimates are not very good since the model does not fit the data well.
Are you ready for the spotlight? We're accepting content ideas for SAS Innovate 2025 to be held May 6-9 in Orlando, FL. The call is open until September 25. Read more here about why you should contribute and what is in it for you!
ANOVA, or Analysis Of Variance, is used to compare the averages or means of two or more populations to better understand how they differ. Watch this tutorial for more.
Find more tutorials on the SAS Users YouTube channel.