Standardized and unstandardized esiti... PreviousNext
Mplus Discussion > Confirmatory Factor Analysis >
Message/Author
 emmanuel bofah posted on Monday, July 29, 2013 - 9:53 am
I have four groupings and doing multigroup analysis. My knowledge with mlpus is not strong enough for me to write a a further syntax to be able to do a pairwise analysis. What I've been doing is to change the reference group in four different analysis :-).
I realized with the four different outputs the unstandardized estimates are the same but the standardized are different(very little). since the standardized estimates are different i am force to use the unstandardized estimates and interprete them but now i have the problem how to interprets them. can i get help on how to write one syntax for this pairwise comparison.I have Mplus license.
secondly and how do i interpret the unstandardized estimates.
 Linda K. Muthen posted on Monday, July 29, 2013 - 1:04 pm
I am not clear on your question.
 emmanuel bofah posted on Monday, July 29, 2013 - 1:35 pm
I am doing multigroup analysis with four groups(gp). i want to a pairwise analysis.that is for instance making gp 1 the reference and compare, then gp 2 the reference group and compare till all four gp are mad reference group in four different analysis.
I realized with the four different outputs the unstandardized estimates are the same but the standardized are different(very little). can i get help on how to write one syntax for this pairwise comparison.
 Linda K. Muthen posted on Monday, July 29, 2013 - 1:41 pm
Please send the outputs and your license number to support@statmodel.com.
 emmanuel bofah posted on Tuesday, July 30, 2013 - 5:58 am
Hi Linda,
One thing i have observe is that if your doing multigroup analysis, the estimates and interpretation is simple with two or three groups. with two groups the reference group dont matter because the factor mean estimates are the same in a reverse order if you change the reference group. but with three groupings the factor mean estimates are in a reverse order ONLY for the unstandardized estimates. but with four groups whenever you change the reference group the mean estimates are different for both standardized and unstandardized estimates. Is it normal? I have send an output for my work on four groups whereby i change the reference group in each model for your attention and advice.
 Linda K. Muthen posted on Tuesday, July 30, 2013 - 10:16 am
When you change the group where the means are zero, the other mean estimates will change. This is expected It will not effect the significance of the means.
 emmanuel bofah posted on Tuesday, July 30, 2013 - 11:15 am
i had a situation where changing the group changed the significance level of the means to non significance. what are the possible cause of that.
 Bengt O. Muthen posted on Tuesday, July 30, 2013 - 12:39 pm
Basic multiple-group theory implies that changing the group changes the scale of the factor means so you should not expect that a test against zero gives the same result. Instead, you should focus on testing the difference between two factor means which can be done using Model Constraint where the difference is a New parameter.
 emmanuel bofah posted on Wednesday, July 31, 2013 - 7:52 am
I n reference to the above discussion. i've tried the following syntax but not working for me.I have 4 groups.
model:
CP BY A8 A10 A11 A12 A15 A19 A22 B3 B4 B8 B10 A25(lam#1-lam#12);
SUC BY B9 B2 B1 B15(lam#13-lam#16);
TECH BY A26 A24 A27 A21 A3 A6(lam#17-lam#22);
FAM BY A23 A17 A18(lam#23-lam#25);
B3 WITH B4;

MODEL CONSTRAINT:
NEW(diff1 diff2 diff3 diff4 diff5);
Do(1,25)diff#=lambda1# lambda4#;

the first error is *** ERROR
Unknown parameter label in MODEL CONSTRAINT: LAMBDA11 which i cannot fix.
will appreciate if i get help in writing this syntax.
 Linda K. Muthen posted on Wednesday, July 31, 2013 - 11:27 am
Example 5.33 shows the correct formulation for this.
 emmanuel bofah posted on Wednesday, July 31, 2013 - 4:37 pm
I have Mplus Base Program and Multilevel Add-On but cannot perform Example 5.33. it is nornal. i get the error "This analysis is only available with the Mixture or Combination Add-On." but Multilevel Add-On is higher than mixture. Is it normal
 Linda K. Muthen posted on Wednesday, July 31, 2013 - 5:04 pm
If you don't have mixture, you can't run the example. You can see how it is set up and copy that for your case.
 emmanuel bofah posted on Wednesday, July 31, 2013 - 6:02 pm
I have try the syntax below but still not working.
USEV ARE
A8 A10 A11 A12 A15 A19 A22 B3 B4 B8 B10 A25 A3 A6 A21 A24 A26 A27
A17 A18 A23 B2 B9 B1 B15 groups ;

Grouping is groups (1=fcoed 2=gsingle 3=bsingle 4=mcoed );

define:
B4=6-B4; B3=6-B3; B10=6-B10;
A11=6-A11; A22=6-A22; A10=6-A10;
A24=6-A24; A27=6-A27; A21=6-A21;

ANALYSIS:

ESTIMATOR = MLR;
PROCESSORS = 2;
ALIGNMENT = FIXED;

model:
F1 BY A8 A10 A11 A12 A15 A19 A22 B3 B4 B8 B10 A25*(lam#_1-lam#_12);
F2 BY B9 B2 B1 B15*(lam#_13-lam#_16);
F2 BY A26 A24 A27 A21 A3 A6*(lam#_17-lam#_22);
F4 BY A23 A17 A18*(lam#_23-lam#_25);

[A8 A10 A11 A12 A15 A19 A22 B3 B4 B8 B10 A25](nu#_1-nu#_12);
A3 A6 A21 A24 A26 A27 A17 A18 A23 B2 B9 B1 B15](nu#_1-nu#_25);

F1-F4@1;
[F1-F4@0];

MODEL PRIORS:
DO(1,25) DIFF(lam1_#-lam4_#)~N(0,0.01);
DO(1,25) DIFF(nu1_#-nu4_#)~N(0,0.01);

PLOT: TYPE = PLOT2;
OUTPUT: TECH1 TECH8 STDYX ALIGNMENT;
 Linda K. Muthen posted on Thursday, August 01, 2013 - 6:03 am
Please send the output and your license number to support@statmodel.com.
 emmanuel bofah posted on Thursday, August 01, 2013 - 7:57 am
Hi Prof Linda,
I have send the output and license hopefully someone reply me this time. This is the second time i've been ask to send the output. No one responded the last time.
 Linda K. Muthen posted on Thursday, August 01, 2013 - 10:17 am
You misspelled CONSTRAINT

CONSTRIANT
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: