Test for skewness and kurtosis PreviousNext
Mplus Discussion > Confirmatory Factor Analysis >
Message/Author
 Anonymous posted on Monday, May 16, 2005 - 1:14 pm
I am conducting CFA with continuous latent variables using Mplus version 3.1. I am using ML to estimate missing data. I have tested for univariate skewness and kurtosis of the measured indicators using SPSS (without estimating missing data). However, I am unsure how well the results of SPSS tests generalize to the distribution of the data that includes estimates. Is there a way for me to test for skewness and kurtosis within Mplus using ML estimation for missing data?

Thanks so much for your assistance.

-Elizabeth
 Linda K. Muthen posted on Monday, May 16, 2005 - 1:27 pm
You can use TECH11 of the OUTPUT command. You would need to use TYPE=MIXTURE MISSING BASIC; with CLASSES = c(1); because TECH11 is available only with MIXTURE.
 Anonymous posted on Tuesday, May 17, 2005 - 9:03 am
Thanks for your prompt and informative reply.

-Elizabeth
 Angela Wolff posted on Saturday, April 12, 2008 - 7:49 am
Hi, I am trying to obtain the TECH13 output to asses multivariate skew and kurtosis. I have cateforical variables and identified 1 class.
DATA:
LISTWISE = ON;

VARIABLE:
CLASSES =C(1)
CATEGORICAL ARE ....;
ANALYSIS:
TYPE = MIXTURE;
OUTPUT:
TECH13;

I try different TYPE but keep getting various messages.

When I used TYPE = MIXTURE MISSING BASIC, I received this message....
TECH13 option is not available for MIXTURE MISSING.
Request for TECH13 is ignored.

When I used, TYPE = MIXTURE I received this message (also LISTWISE DELETION IS ON)...
TECH13 option is only available for models with y-variables.
Request for TECH13 is ignored.

Can you assist please in obtaining this output. Thanks
 Linda K. Muthen posted on Saturday, April 12, 2008 - 8:26 am
Please send your input,data, output, and license number to support@statmodel.com.
 Roisin O'Connor posted on Wednesday, May 21, 2008 - 7:55 am
I am trying to test the multivariate normality for a CFA analysis (23 items, 4 factors). I see from the code above that to do this I should use the mixture option and identify one class. When I do this I get the following error:

TECH11 option is not available for single-class models.
Request for TECH11 is ignored.

Any thoughts?
 Roisin O'Connor posted on Wednesday, May 21, 2008 - 9:43 am
To add to my previous question, I should clarify that I want the multivariate normality statistics for data that includes missing values.

Thanks.
 Linda K. Muthen posted on Wednesday, May 21, 2008 - 11:15 am
I think you mean to be using TECH13.

In Version 5, TYPE=MISSING is the default.
 JOEL WONG posted on Thursday, May 09, 2013 - 1:05 pm
I am trying to test for multivariate normality and I got this message:
"TECH13 option is not available for MIXTURE MISSING."

Under variables, I entered
classes = c (1);

and in my model, I indicated
%OVERALL%

I also entered TECH13 under output.

It works fine for TECH12 but not TECH13.

Could you help me figure out the problem?
 Linda K. Muthen posted on Thursday, May 09, 2013 - 1:36 pm
You have the Base program. You need to Base program and the Mixture add-on.
 JOEL WONG posted on Thursday, May 09, 2013 - 1:51 pm
Thanks for responding so quickly.

Based on an old forum post by Bengt, I figured that TECH12 also produces an output to test for multivariate normality.

I could run the output for TECH12 but I am not sure if I understand the output. Do I look at "Observed skewness" and "Observed Kurtosis" to examine multivariate skewness/kurtosis or do these refer to univariate skewness/kurtosis?

Thank you so much.
 Linda K. Muthen posted on Thursday, May 09, 2013 - 2:24 pm
TECH12 is only available for TYPE=MIXTRE.
 Luisa Solms posted on Thursday, October 08, 2020 - 12:17 am
Hi Linda and team,

I have the same question as Roisin O'Connor. I want to conduct a CFA with 5 latent factors and 41 items and I would like to obtain the multivariate normality statistics for the data. The data includes missing values.

I have tried the TECH11, TECH12 and TECH13 output but nothing seems to work.

My syntax is the following:

VARIABLE:
NAMES ARE Eng_1 Eng_2 Eng_3 Eng_4 Eng_5 Eng_6 Eng_7 Eng_8 Eng_9 WP_1 WP_2 WP_3 WP_4 WP_5 WP_6 WP_7 WP_8 WP_9 WP_10 WP_11 WP_12 WP_13 BO_1 BO_2 BO_3 BO_4 BO_5 BO_6 BO_7 Pos_1 Pos_2 Pos_3 Pos_4 Pos_5 Pos_6 Neg_1 Neg_2 Neg_3 Neg_4 Neg_5 Neg_6;
USEVARIABLES ARE Eng_1 Eng_2 Eng_3 Eng_4 Eng_5 Eng_6 Eng_7 Eng_8 Eng_9 WP_1 WP_2 WP_3 WP_4 WP_5 WP_6 WP_7 WP_8 WP_9 WP_10 WP_11 WP_12 WP_13 BO_1 BO_2 BO_3 BO_4 BO_5 BO_6 BO_7 Pos_1 Pos_2 Pos_3 Pos_4 Pos_5 Pos_6 Neg_1 Neg_2 Neg_3 Neg_4 Neg_5 Neg_6;

MISSING ARE ALL (-999);

CLASSES = C(1);

ANALYSIS:
TYPE IS MIXTURE;

MODEL:
%OVERALL%
F1 by Eng_1-Eng_9 ;
F2 by WP_1-WP_9;
F3 by BO_1-BO_7;
F4 by Pos_1-Pos_6;
F5 by Neg_1-Neg_6;

OUTPUT: TECH11;
 Luisa Solms posted on Thursday, October 08, 2020 - 12:29 am
It worked when using the TECH12 output! Just one other question, the Mplus output also contains values for estimated mixed kurtosis and residuals for mixed kurtosis. However, there are only zero's for the estimated mixed kurtosis. Can I only interpret the observed kurtosis values or do I also have to interpret the estimated values and residuals?
 Bengt O. Muthen posted on Thursday, October 08, 2020 - 11:24 am
Send your output to Support along with your license number.
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: