Wald chi-square test PreviousNext
Mplus Discussion > Multilevel Data/Complex Sample >
Message/Author
 fangfang posted on Wednesday, September 16, 2015 - 12:25 am
Dear Dr. Muthen,

I want to use the Wald chi-square test to test whether there are differences in the coefficients c1 and c2. I wrote a set of codes according the user guide. I want to know which code is the right one?
It would be great if you could provide an appropriate example input code. I would be grateful for your help! Thanks in advance.
The first,

MODEL CONSTRAINT:
New (c);
MODEL TEST:
c=c1-c2;

the second,
MODEL CONSTRAINT:
New (c);
c=c1-c2;
MODEL TEST:
c=0;
 Linda K. Muthen posted on Wednesday, September 16, 2015 - 10:34 am
You use the labels from the MODEL command in MODEL CONSTRAINT and MODEL TEST. Neither is correct.

MODEL:
y ON x;
MODEL male:
y ON x (p1);
MODEL female:
y ON x (p2);

MODEL TEST:
0 = p1 - p2;
 fangfang posted on Wednesday, September 16, 2015 - 6:01 pm
Thank you very much!Dr. Muthen.
I am sorry for not making it clear.
The model I want to run is the regression of two dependent variables (Y1, Y2) on one independent variable (X).

MODEL:
y1 ON x(p1);
y2 ON x(p2);
MODEL TEST:
0=p1-p2;

is that right? thank you again!
 Linda K. Muthen posted on Thursday, September 17, 2015 - 6:10 am
The language is correct but you should do this only if y1 and y2 are measured on the same scale.
 fangfang posted on Thursday, September 17, 2015 - 7:14 am
Thank you for your prompt response! Dr.Muthen.
In the model, y1 and y2 are two different types of one behavior (promotive voice and prohibitive voice). I don't understand what does "the same scale" mean. Do you mean that they are the same behavior measured in different samples?
what should I do if I want to compare the cofficients of the regression of y1,y2 on x in the same sample? could you provide the code, please? Thank you very much!
 Linda K. Muthen posted on Thursday, September 17, 2015 - 9:16 am
Are the minimum and maximum values of the variables the same. If not, they should not be compared. You should look at their standardized coefficients.
 fangfang posted on Thursday, September 17, 2015 - 7:31 pm
Many thanks to you.Dr. Muthen.

It is a test of the difference between coefficients for different dependent variables from a single sample. But, all items of variables are measured by a 7-point Likert-type scale. Given that,is it suitable to use the wald test?

I am grategul for your help!
 Linda K. Muthen posted on Thursday, September 17, 2015 - 8:28 pm
If both dependent variables are measured on the same scale, this is OK.
 fangfang posted on Thursday, September 17, 2015 - 9:16 pm
Thank you very much for your help!Dr. Muthen.
 Hon Chung Choi posted on Tuesday, October 03, 2017 - 10:03 pm
Hi Dr. Muthen,

I'd like to use Wald Chi-square test to assess parameter quality using Bayesian estimator. However, Mplus 8 said it is impossible to do Wald with Bayes.

Are there any alternatives?

Thanks for your time
Tommy
 Hon Chung Choi posted on Tuesday, October 03, 2017 - 10:09 pm
Regarding the above post, here is the output from Mplus,

MODEL TEST is not available with ESTIMATOR=BAYES.
 Bengt O. Muthen posted on Wednesday, October 04, 2017 - 3:17 pm
You can use Model Constraint to define a new parameter that is e.g. the difference between two parameters, so that gives you a test. But you can't test several such differences jointly.

Model Test for Bayes is on our to-do list.
 Freya Glendinning posted on Friday, May 25, 2018 - 10:28 am
Hi,

I am trying to compare two beta coeffients from the same model. I think I am missing a piece of the puzzle from the above...

this is my input which is probably quite wrong:

VARIABLE: NAMES ARE SEX IDENTITY MI SI LI SE DEP RAI;
USEVARIABLES SE IDENTITY DEP MI SI LI;
ANALYSIS: ESTIMATOR = ML;


MODEL CONSTRAINT:
NEW (misi);
WALD = MI - SI;
MODEL:
WB by SE DEP;
WB ON IDENTITY MI;
WB ON IDENTITY SI;
WB ON IDENTITY LI;
MODEL misi:
WB ON MI (P1);
WB ON SI (P2);
MODEL TEST:
0 = P1-P2;

*** ERROR in MODEL command
Unknown group name MISI specified in group-specific MODEL command.

Thanks
 Bengt O. Muthen posted on Friday, May 25, 2018 - 12:42 pm
See the UG index under Wald test.
 Alexia Carrizales  posted on Wednesday, June 27, 2018 - 8:28 am
Hi
I'm testing a cross lagged panel model, I can not figure out what it is wrong with my syntax, the paths from pbs to bes are not significant so we are testing this model across gender
I got a message THE STANDARD ERRORS OF THE MODEL PARAMETER ESTIMATES COULD NOT BE COMPUTED. THE MODEL MAY NOT BE IDENTIFIED. CHECK YOUR MODEL.
PROBLEM INVOLVING THE FOLLOWING PARAMETER: Parameter 106, Group FEMALE: [ PBW1 ]
THE CONDITION NUMBER IS -0.403D-10.
THE ROBUST CHI-SQUARE COULD NOT BE COMPUTED.

PbW3 on PbW2 EmW2;
PbW2 on PbW1 emW1;
EmW3 on EmW2;
EmW2 on EmW1;

PbW1 with EmW1;
PbW2 with EmW2;
PbW3 with EmW3;

PbW1;
PbW2;
PbW3;
EmW1;
EmW2;
EmW3;
! correlated residual errors over time
HLPw1 with HLPw2 HLPw3;
HLPw2 with HLPw3;
CARw1 with CARw2 CARw3;
CARw2 with CARw3;
COGEMw1 with COGEMw2 COGEMw3;
COGEMw2 with COGEMw3;
EMODISw1 with EMODISw2 EMODISw3;
EMODISw2 with EMODISw3;
EMOCONw1 with EMOCONw2 EMOCONw3;
EMOCONw2 with EMOCONw3;
Model male: ! the only difference with the speficied model is this and i did the same for the females
[PbW1];
[PbW2];
[PbW3];
[EmW1];
[EmW2];
[EmW3];
 Bengt O. Muthen posted on Wednesday, June 27, 2018 - 4:13 pm
Send your output to Support along with your license number.
 Snigdha Dutta posted on Tuesday, June 16, 2020 - 3:47 pm
The Wald estimate for male and female group for a particular path was significant. However, the path for both male and female models were non significant. How is this interpreteD?
 Bengt O. Muthen posted on Tuesday, June 16, 2020 - 5:44 pm
Perhaps your first sentence refers to a difference between the male and female estimate in a joint analysis of the genders and your second sentence refers to separate gender analyses and testing against zero?
 Snigdha Dutta posted on Tuesday, June 16, 2020 - 6:24 pm
MODEL BOY:
Y ON X (BOYA);

MODEL GIRL:
Y ON X (GIRLA);

MODEL TEST:
0 = BOYA - GIRLA;

This is the part of the syntax I'm referring to in my question.

The model test produces a significant result suggesting that there is a difference in the predictor-mediator path.

However, these paths for male model and female model are not significant themselves.

Is such an outcome possible?
 Bengt O. Muthen posted on Thursday, June 18, 2020 - 4:20 pm
Yes this is possible. Think of this case:

male est = -0.5

female est = +0.5

The distance between each of those estimates and zero is smaller (0.5) than the distance between the estimates (1.0).
 Snigdha Dutta posted on Thursday, June 18, 2020 - 6:39 pm
I see. In this case, would I still interpret this as significant differences between groups? Even though the parameters for male and female individually are not significant?
 Bengt O. Muthen posted on Friday, June 19, 2020 - 6:12 pm
Yes on both.
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: