Significance test in LCM PreviousNext
Mplus Discussion > Growth Modeling of Longitudinal Data >
Message/Author
 Shige Song posted on Saturday, August 26, 2006 - 7:10 pm
I have a general question: how many ways are there in Mplus to test significance of fixed and random parameters in LCM, and which one is generally recommended?

I have a substantively interesting but statistically insignificant parameter (based on Est./S.E.) because the standared error is too big, should I throw this parameter away?

By the way, I am estimating a two-level LCM, and the parameter of interest is at the community-level (level-2).

Thanks!

Shige
 Bengt O. Muthen posted on Sunday, August 27, 2006 - 6:12 pm
Well, let's see - we have z scores, bootstrapped CIs, Wald test (Model Test), and loglikelihood difference (chi-square) testing.

I would not necessarily fix an insignificant parameter, but report it as such. Perhaps the sample size is not sufficient, which sometimes happens on level 2. But sometimes the data doesn't carry enough information about a parameter in the specified model.
 Shige Song posted on Thursday, August 31, 2006 - 3:06 am
Hi Bengt,

That makes a great deal of sense, thanks!

Shige
Back to top
Add Your Message Here
Post:
Username: Posting Information:
This is a private posting area. Only registered users and moderators may post messages here.
Password:
Options: Enable HTML code in message
Automatically activate URLs in message
Action: