p-values from summary and mxCompare do not match

2 replies [Last post]
JWiley's picture
Offline
Joined: 03/25/2011

Specifically, I fit a model as well as a saturated model. When calling summary, I pass the -2LL and DF from the saturated model, with mxCompare, I compare my model with the saturated. Here are the results, everything else is as expected. Am I missing something?
from summary():
chi-square: 5.35
p: 1

from mxCompare():
minus2LL df AIC diffLL diffdf p
1 5318 742 3834 NA NA NA
2 5323 744 3835 5.35 2 0.0691

A reproducible example is provided in the attached script. I am using r57586 on Windows x64 with OpenMx_999.0.0-1661.

AttachmentSize
pvalExample.R1.82 KB
mspiegel's picture
Offline
Joined: 07/31/2009
mxCompare() invokes the

mxCompare() invokes the summary() function on the 'base' argument, and then invokes the summary() function on the 'comparison' argument, but in either case it does not pass any of the optional arguments (SaturatedLikelihood, SaturatedDoF, IndependenceLikelihood, IndependenceDoF) to the summary() function.

Whether or not mxCompare() should be passing optional arguments to the summary() function is a question for the other people on the development team.

tbates's picture
Offline
Joined: 07/31/2009
support mxCompare(numbs, numStats, SaturatedLikelihood, indep)

Makes sense to me that numbs, numStats, SaturatedLikelihood, and indep should be passed on to the omxLapply calls to summary if they are provided to mxCompare()