Appropriate estimator PreviousNext
Mplus Discussion > Categorical Data Modeling >
Message/Author
 Anonymous posted on Monday, November 12, 2001 - 9:41 pm
Which estimator in Mplus would you recommend for dealing with categorical dependent variables, nonnormality (highly skewed), AND incomplete data? Can any estimator deal with the three problems simultaneously?

If not, can any estimator deal with categorical dependent variables AND incomplete data?

Thank you very much in advance!
 Linda K. Muthen posted on Tuesday, November 13, 2001 - 8:02 am
Mplus cannot deal with categorical and incomplete together. Until Mplus has missing with categorical, we recommend imputing several datasets using, for example, Joe Schafer's program, and then using the RUNALL facility to analyze the data. For categorical and nonnormality, I would recommend WLSMV.
 Anonymous posted on Tuesday, November 13, 2001 - 12:15 pm
Thank you very much for the answer. Then, which estimator in Mplus would you recommend for dealing with nonnormality (highly skewed) AND incomplete data?
 Linda K. Muthen posted on Wednesday, November 14, 2001 - 6:59 am
I can't recommed an estimator for categorical outcomes and missing as Mplus does not handle this combination. If you mean with the datasets imputed outside of Mplus as mentioned above, I would always recommend WLSMV for skewed categorical data. If you mean for continuous outcomes and missing data in Mplus, the only estimator available is ML. Mixture modeling in Mplus also handles missing and sometimes skewed data can be seen as a combination of two normal distribtuions coming from different populations. Here MLR would be recommended.
 Julie Hall posted on Thursday, May 11, 2006 - 12:25 pm
Hello,
I am a new Mplus user, and have run into a problem when running the following model (see input below). One of my DVs is categorical and skewed, so I am trying to use WLSMV. However, no model fit indices (or errors/warnings) appear in the output. But if I use MLR with montecarlo integration then the full output appears. Why might this happen and which estimator should I use? Thank you in advance!

DATA:
FILE IS "C:\Documents and Settings\Julie\My Documents\Dissertation\scored data for
diss analyses - no raw data\diss data without breakups.dat";
FORMAT IS 114F8.2;

VARIABLE:
NAMES ARE id gender1 age1 year1 relaffi1 ethnic1 semester length1
datexc1 datnon1 engag1 marr1 livsep1 livtog1 longdis1 kiss1 hug1 pet1
oral1 interc1 date1 romfeel1 other1 infnum1 cheat1o sphyinf1 semoinf1
cheat1s pphyinf1 pemoinf1 cheat1p cheat1sp chea1spo inftype1 pcheatp1
pcheate1 prqctot1 kmstot1 rastot1 rsstot1 depress1 staitot1 baitot1
shame1 guilt1 extrtot agretot constot neurtot opentot risctot
justtot1 continf1 gwstot1 dropout gender2 age2 year2 relaffi2 ethnic2
romrel2 samerel2 stilltog length2 datexc2 datnon2 engag2 marr2
livsep2 livtog2 longdis2 breakup kiss2 hug2 pet2 oral2 interc2 date2
romfeel2 other2 infnum2 cheat2o sphyinf2 semoinf2 cheat2s pphyinf2
pemoinf2 cheat2p cheat2sp chea2spo inftype2 hall16 severc2 peverp2
pevere2 depress2 staitot2 baitot2 gwstot2 shame2 guilt2 prqctot2
kmstot2 rastot2 rsstot2 justtot2 continf2;
USEVARIABLES chea1spo prqctot1 kmstot1 rastot1 rsstot1 depress1
staitot1 shame1 guilt1 gwstot1 chea2spo prqctot2 kmstot2 rastot2 rsstot2
depress2 staitot2 gwstot2 shame2 guilt2;
CATEGORICAL chea2spo;

MISSING IS BLANK;

ANALYSIS:
TYPE IS MISSING;
ESTIMATOR IS wlsmv;

MODEL:
dist1 by depress1 staitot1 shame1 guilt1 gwstot1;
dist2 by depress2 staitot2 gwstot2 shame2 guilt2;
relqual1 by prqctot1 kmstot1 rastot1 rsstot1;
relqual2 by prqctot2 kmstot2 rastot2 rsstot2;
depress1 with depress2;
staitot1 with staitot2;
shame1 with shame2;
guilt1 with guilt2;
gwstot1 with gwstot2;
prqctot1 with prqctot2;
kmstot1 with kmstot2;
rastot1 with rastot2;
rsstot1 with rsstot2;
dist1 with relqual1;
chea1spo with dist1;
chea1spo with relqual1;
dist2 on dist1;
relqual2 on relqual1;
relqual2 on dist1;
dist2 on relqual1;
chea2spo on chea1spo;
dist2 on chea1spo;
relqual2 on chea1spo;
chea2spo on dist1;
chea2spo on relqual1;


OUTPUT:
SAMPSTAT;
MODINDICES(5);
STANDARDIZED;
CINTERVAL;
 Linda K. Muthen posted on Thursday, May 11, 2006 - 12:27 pm
This is a support question. Please send your input, data, output, and license number to support@statmodel.com.
 f f  posted on Thursday, May 15, 2014 - 7:01 pm
Hello,

I need a reference to justify why I should use WLSMV estimator to deal with ordinal skewed data.

I would be grateful if introduce some references.

Regards,
 Linda K. Muthen posted on Friday, May 16, 2014 - 10:01 am
Muthén, B., Kaplan, (1985). A comparison of some methodologies for the factor analysis of non-normal Likert variables. British Journal of Mathematical and Statistical Psychology, 38, 171-189.
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: