Difftest in path to categorical DV PreviousNext
Mplus Discussion > Categorical Data Modeling >
Message/Author
 Mary Beth Oliver posted on Friday, March 05, 2010 - 1:42 pm
Hello All,

I'm trying to test for the invariance of a path that predicts a dichotomous DV (ENDLINK). I'm hoping that someone could confirm (or not!) my interpretation. I used the DIFFTEST option.

First, I ran:
USEVARIABLES ARE story comps temp bitotal endlink;
CATEGORICAL IS endlink;
GROUPING IS target (1 = immigrants 2 = elderly 3 = prisoners);
MODEL:
comps on story (1);
temp on comps (2);
bitotal on temp (3);
endlink on bitotal (4);
Model elderly:
endlink on bitotal ;
Model prisoners:
endlink on bitotal ;
SAVEDATA: DIFFTEST IS deriv.dat;

Then I ran:
CATEGORICAL IS endlink;
GROUPING IS target (1 = immigrants 2 = elderly 3 = prisoners);
ANALYSIS: DIFFTEST IS deriv.dat;
MODEL: comps on story (1);
temp on comps (2);
bitotal on temp (3);
endlink on bitotal (4);

And in the output I see:
Chi-Square Test for Difference Testing Value 2.915
Degrees of Freedom 2**
P-Value 0.2328

I just want to make sure that I can interpret this as suggesting that the path from bitotal to enlink does not differ between the three groups.

Many thanks, in advance, for any insight to this new user of MPlus.

Kind regards,

Mary
 Linda K. Muthen posted on Friday, March 05, 2010 - 10:00 pm
That interpretation is correct. Note that the default in Mplus is that regression coefficients are not held equal across groups. You could have specified the first model as:

MODEL:
comps on story (1);
temp on comps (2);
bitotal on temp (3);
endlink on bitotal ;
 Mary Beth Oliver posted on Saturday, March 06, 2010 - 1:26 am
Thank you so very much for your help! This is great news. Fantastic!

With kind regards,

Mary
 LiChong posted on Sunday, May 02, 2010 - 10:41 am
Hello, Dr. Muthen --
about the DIFFTEST,i read the paper "robust chi square difference testing with mean and variance adjusted test statistics",the paper studied the type 1 rates, both rates are sufficiently close to the nominal rejection rate of 5%.
then, i now want to know the power of the DIFFTEST, for example if the measurement invariance don not exist, can DIFFTEST successfully detecte it ?if there are 500 such situations, how many the DIFFTEST can successfully explore the measurement invariace.
i cannot find the paper exploring this, can you give me some adivice?
thank you very much for your any advice and help.
 Linda K. Muthen posted on Sunday, May 02, 2010 - 2:20 pm
I don't know of any paper that studied power for DIFFTEST. There is a DIFFTEST technical appendix on the website that also shows the proper way to study DIFFTEST giving example inputs for the two steps needed.
 LiChong posted on Monday, May 03, 2010 - 2:17 am
thank you.
i think it is necessary to study power of DIFFTEST,and i will try doing this thing.
thanks again for your advice.
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: