Testing for significants PreviousNext
Mplus Discussion > Multilevel Data/Complex Sample >
 victor china posted on Wednesday, July 25, 2007 - 1:17 pm

I would like to know how i test for my models significants?

my model is a two level path analysis (with both binary and continuous variables), what is the appropriate null measure to test against for model selection?

the structure of my model is:
CATEGORICAL = fin ini tact pose;
tact ON logTL ;
time ON logTL;
ini ON logTL ;
pose fin ON tact ;
time ON pose ini fin ;
tact fin ON ini ;
fin ON pose ;

tact time ON N ;

thanks in advance
 Linda K. Muthen posted on Wednesday, July 25, 2007 - 2:20 pm
An alternative model would be one that includes some of the paths on the within level that are not in the model above.
 victor china posted on Thursday, July 26, 2007 - 5:06 am
thank you,

are you saying that in general there is no formal null model provided by mplus (that includes for example all the possible relationships between variables)?

also if i understand correctly my model should be nested within the null model? and not the other way around?

 Linda K. Muthen posted on Thursday, July 26, 2007 - 2:26 pm
We refer to the H0 model as the null model. The H1 model is a totally saturated model.
 victor china posted on Friday, July 27, 2007 - 6:54 pm

one more final technical question,

for calculating chi-square difference test based on loglikelihood values as explained in your site, i need to specify how many parameters my models have.

do i use the value defined in the output?

"Information Criteria

Number of Free Parameters 22"

if not then how can i calculate the amount of parameters?

thanks in advance.
 Linda K. Muthen posted on Friday, July 27, 2007 - 7:44 pm
The number of free parameters is given in the output as you show.
Back to top
Add Your Message Here
Username: Posting Information:
This is a private posting area. Only registered users and moderators may post messages here.
Options: Enable HTML code in message
Automatically activate URLs in message