proc fmm

n6 — Mon, 18 Nov 2024 19:07:47 GMT

I'm doing a Zero-Truncated Negative Binomial model in proc fmm. (I'm also considering a Zero-Truncated Poisson although I think we have overdispersion.)

One general question is this. From what I read diving the Pearson Statistic by the degrees of freedom tells you whether there is overdispersion or underdispersion. A ways < 1 implies underdispersion and a ways > 1 implies overdispersion.

Okay, that's one thing. But more specifically, in one model I get the following message in the output:

"Optimization routine cannot improve the function value."

And this same model does not give me results. Any idea what this is? It kinda feels like the "Algorithm did not converge" message I sometimes get in other SAS procs.

Any info is appreciated. Thanks in advance.

Re: proc fmm

SAS_Rob — Tue, 19 Nov 2024 15:03:42 GMT

Yes, you could test for overdispersion by using the Pearson/DF.
The second message is an indication that the model did not converge. A suggestion would be to simplify the model and add complexity back from there.

topic proc fmm in Statistical Procedures

proc fmm

Re: proc fmm