large number of models

[megpetrie]

Hello,
I have a large number of models that I am trying to test. After inputting all the formulas, I used the following code to execute the mark model analysis:

Code: Select all

sh.model.list=create.model.list("CJS")
sh.wrap=mark.wrapper(sh.model.list, data=sh.process, ddl=sh.ddl, adjust=T, se=T, invisible=F, realvcv=T, profile.int=T, chat=1.302)
sh.wrap


The sh.wrap call only returned a portion of the models. Is there a limit to the number of models that can be returned in a model table? How can I just return the say, 50 models with the lowest AIC?

Thanks so much,
Megan

[jlaake]

Megan-

Only 2 reasons for a model to not appear. First was that the input file was not constructed. I suggest that you add run=F to see if you have a problem with some of the model specifications. Second is the output file was not constructed completely. Maybe some are not converging but you usually get a complete output file. You can always run the models separately with mark to identify the issue. You may have to look at the output file outside of R.

To trim the models, you'll have to do that after the fact. See remove.mark. You can get the model numbers from the model.table that have delta aic larger than what you want and specify those to remove.mark which creates and returns a new marklist without the specified models.

--jeff