Message/Author |
|
Anonymous posted on Wednesday, January 25, 2012 - 12:46 am
|
|
|
I have been using the knownclass option for the analysis of categorical data in multiple groups in conjunction with MLR and to date this works fine. Is it possible to perform these analyses using the grouping option and a different estimator? Is it possible to perform multiple-group latent class analyses in Mplus of categorical data? Both would require the classes command if the knownclass option is used and I assume this is not possible. |
|
|
Not knowing more about your model, I assume you have categorical factor indicators. In this case, you need the KNOWNCLASS option for grouping. If you use WLSMV, you can use the GROUPING option. There is no difference between using GROUPING and KNOWNCLASS. |
|
Anonymous posted on Thursday, January 26, 2012 - 2:54 pm
|
|
|
I am thinking about how to specify a factor mixed model analysis with binary/categorical indicators. If I used WLSMV as an estimator could I use GROUPS to specify the groups and CLASSES to specify the number of latent classes? |
|
|
If you have latent classes, you need TYPE=MIXTURE. Only maximum likelihood using the KNOWNCLASS option is available. |
|
Anonymous posted on Monday, January 30, 2012 - 2:05 am
|
|
|
Is it possible to perform multiple group latent class analyses in Mplus? |
|
|
Yes. You would use the KNOWNCLASS option. |
|
|
Hi I'm trying to test for group differences in a discrete-time survival analyses; however, I keep getting an error that doesn't make sense. I've included my syntax and output below. categorical are SEX13 SEX14 SEX15 SEX16 SEX17; dsurvival are SEX13 SEX14 SEX15 SEX16 SEX17; missing are all(-9999); CLUSTER is CLTSCID; class = sex (2); knownclass = sex (GIRL = 0 GIRL = 1); ANALYSIS: Type = Mixture Complex; ESTIMATOR=MLR; ALGORITHM=INTEGRATION; PROCESSORS=2; MODEL: %OVERALL% SEXDBT by SEX13@1 SEX14@1 SEX15@1 SEX16@1 SEX17@1; SEXDBT on PRTKNW SLCINF YTHDSC CNTRL1 FAMRLE MTLRLT USPV35 BLACK LATINO OTHER ; SEXDBT@0; MODEL sex: %sex#1% SEXDBT on PRTKNW; %sex#0% SEXDBT on PRTKNW; OUTPUT: sampstat STDYX; PLOT: TYPE IS PLOT3 *** ERROR in MODEL command Unknown class model name SEX specified in C-specific MODEL command. |
|
|
The sex classes are numbered 1, 2, not 1, 0. |
|
|
Doh! Thank you! |
|
|
Thank you for your time and patience. I've updated my syntax but I'm still receiving the same error. Here is the updated syntax. VARIABLE: names are id PRTKNW SLCINF USPV35 USPV57 NOSPV YTHDSC MTLRLT SEX13 SEX14 SEX15 SEX16 SEX17 CENSOR CLTSCID GIRL Black Latino OTHER CNTRL1 FAMRLE PRTMTR ; usevariables are PRTKNW SLCINF YTHDSC CNTRL1 FAMRLE MTLRLT USPV35 BLACK LATINO OTHER SEX13 SEX14 SEX15 SEX16 SEX17; categorical are SEX13 SEX14 SEX15 SEX16 SEX17; dsurvival are SEX13 SEX14 SEX15 SEX16 SEX17; missing are all(-9999); CLUSTER is CLTSCID; classes = cg (2); knownclass = cg (GIRL = 0 GIRL = 1); ANALYSIS: Type = Mixture Complex; ESTIMATOR=MLR; ALGORITHM=INTEGRATION; PROCESSORS=2; MODEL: %OVERALL% SEXDBT by SEX13@1 SEX14@1 SEX15@1 SEX16@1 SEX17@1; SEXDBT on PRTKNW SLCINF YTHDSC CNTRL1 FAMRLE MTLRLT USPV35 BLACK LATINO OTHER ; SEXDBT@0; MODEL cg: %cg#1% SEXDBT on PRTKNW; %cg#2% SEXDBT on PRTKNW; OUTPUT: sampstat STDYX; PLOT: TYPE IS PLOT3 *** ERROR in MODEL command Unknown class model name CG specified in C-specific MODEL command. |
|
|
Remove MODEL cg. You don't need this with only one categorical latent variable. |
|
Back to top |