Build a Mplus data file, command file, execute the command file and display Mplus log file in Stata results window. Parameter estimates, standard errors, and model and fit statistics are stored as saved results. runmplus is a modification and extension of Michael Mitchell's stata2mplus module (see http://www.ats.ucla.edu/stat/Stata/). The Mplus demo version and new version 6 Bayesian estimator options are supported.
-------------------------------------------------------------------------------
runmplus varlist [if] [in] , [ SAVEINPutfile(string) SAVELOGfile(string) saveinputdatafile(string) log(string) TItle(string) DATa(string) VARiable(string) DEFine(string) ANalysis(string) MOdel(string) OUTput(string) savedata(string) PLOT(string) mc(string) montecarlo(string) population(string) TItle(string) demo6 Mplus_command_option_shortcuts Special_postestimation_commands_options ]
Mplus command option shortcuts:
missing(#) IDvariable(string) NGroups(string) CATegorical(string) GROUPing(string) weight(string) CLUSter(string) within(string) between(string) tscores(string) centering(string) ttype(string)
type(string) ESTimator(string) MATrix(string) PARAMeterization(string) VARiance(string) ITERations(string) SDITERations(string) H1ITERations(string) MITERations(string) MCITERations(string) MUITERations(string) CONvergence(string) H1CONvergence(string) COVERage(string) logcriterion(string) MCONVergence(string) MCCONVergence(string) MUCONVergence(string) mixc(string) mixu(string) LOGHigh(string) LOGLow(string) UCELLsize(string) ALGOrithm(string)
SAMPStat MODindices(string) STANDardized RESidual CINTerval NOCHIsquare h1se H1TEch3 PATterns FSCOEFficeint FSDETerminacy tech(string) population(string) loadresults varnocheck
Required commands
varlist - Variables to be used in the Mplus model. This command is not actually required. If it's left blank runmplus will assume you want all of the variables, unless there is text in the MC or montecarlo commands. NOTE: This varlist can be a STATA abbreviated varlist. Keep in mind that Mplus and STATA expand varlists differently, and what works in STATA may not work in Mplus. For example, the abbreviated list y1-x3 will identify all variables falling between and including y1 and x3, and will work as a varlist in this command. However, you couldn't use the option categorical y1-x3 because Mplus can't expand such a varlist.
Options
demo6 - If you have both the demo version 6 and another version of Mplus, demo6 will force a call to the Mplus demo version 6. If you only have the demo version, you do NOT need to specify this option. runplus will find the right executable so long as they are in the default install locations. Specify "demo6" if running the demo version 6 of Mplus. No other demo versions are supported but you could edit the ado to achieve that (c. line 600).
missing - Specify number for missing values, default -9999
SAVEINPutfile(string) ,
SAVELOGfile(string) and
saveinputdatafile(string) - runplus uses temporary file names to store the Mplus command file and data file. By default, these are not saved. If the user wishes to have these files to review after execution of runplus, file names can be specified using the three save options. DOS shell copy commands are used and replacement of existing files will occur. NOTE: don't include the suffexes (inp, out, dat).
log(string) - if "off" is specified, runmplus supresses the dump of the Mplus output file to the STATA results window. Any other entry will have no influence on runmplus processing. Can be used at the same time as savelogfile.
Mplus_command_file_options - See below and the Mplus Users Guide. The Mplus command file options are captured as strings and placed into a ascii file that is sent to the Mplus program. You can put anything you want in any of these options, but only options that make sense to Mplus will be useful. Modifications or special circumstances are addressed in the next section.
Description
runmplus brings the powerful latent variable analysis capabilities of Mplus (see www.statmodel.com) into the Stata enviroment. While runmplus does little more than data reformat and ascii file creation, it removes a lot of the hassle of estimating categorical and continuous latent variable models using Mplus. runmplus is a modification and extension of Michael Mitchell's stata2mplus module (see http://www.ats.ucla.edu/stat/Stata/). The runmplus module does a number of tasks:
(1) creates an ascii data file suitable for use with Mplus using code from stata2mplus, and also
(2) creates a mplus command file using options specified,
(3) executes the mplus command file using the dos shell command feature, and
(4) dumps the Mplus model results to the screen/log (unless directed not to by the user using log(off)).
Mplus is required to use this STATA module. If you don't have Mplus, visit www.statmodel.com. Moreover, runmplus assumes mplus.exe is located in the windows path (or that mpdemo6.exe is in c:\program files\mplus demo\). runmplus will not check code for errors. runmplus can't be used efficiently without an understanding of the Mplus command syntax.
Known Bugs and Limitations
Long variable names: Mplus does not support long variable names, so neither does runmplus. Your runmplus/Mplus file will run with a long file name so long as the first eight characters are unique. It'd be better to make them short in STATA first.
runmplus will not allow you to use the PLOT option in Mplus.
runmplus may have trouble reading parameter estimates when more than one latent class variable is used. Use the saveinp and saveinputdatafile options and run the model from Mplus.
Examples -------------------------------------------------------------------------------
Simple linear regression for a continuous dependent variable
. runmplus y1 x1 x3, model(y1 on x1 x3;)
Logistic regression
. runmplus y1 x1 x3, categorical(y1) type(logistic) model(y1 on x1 x3;)
Path analysis with continuous dependent variables
. runmplus y1-y3 x1-x3, model(y1 y2 on x1 x2 x3; y3 on y1 y2 x2;)
Path analysis with categorical dependent variables
. runmplus y1-y3 x1-x3, categorical(y1 y2 y3) model(y1 y2 on x1 x2 x3; y3 on y1 y2 x2;)
Path analysis with a combination of continuous and categorical dependent variab > les
. runmplus y1-y3 x1-x3, categorical(y3) model(y1 y2 on x1 x2 x3; y3 on y1 y2 x2;)
Exploratory factor analysis with continuous indicators
. runmplus y1-y12, type(efa 1 4)
Exploratory factor analysis with categorical indicators
. runmplus y1-y12, type(efa 1 4) categorical(all)
Exploratory factor analysis with a mixture of categorical and continuous indica > tors
. runmplus y1-y12, type(efa 1 4) categorical(y1 y3 y5 y7 y9 y11)
Confirmatory factor analysis with continuous indicators
. runmplus y1-y6, model(f1 by y1-y3; f2 by y4-y6;)
Confirmatory factor analysis with continuous indicators and equality constraint > s, using savedata and savelogfile options, and following up with runmplus_load_savedata bringing the factor estimates back into the current data set
. runmplus y1-y6 id , idvariable(id) ///
model(f by y1-y6 *1 (1); f@1; [y1-y6 *0] ; y1-y6 *1 (2) ;) ///
savedata(save=fscores; file=c:\trash\trash.dat) ///
savelogfile(c:\trash\trash) // note no suffix on trash.out
. preserve
. runmplus_load_savedata , out(c:/trash/trash.out) clear // note with suffix on trash.out
. id f
. sort id
. tempfile f1
. save `f1'
. restore
. merge id using `f1' , sort
. table _merge
. drop _merge
Mplus Program Options ------------------------------------------------------------------------------- Data Options not usually needed - The user should not generally need to specify Mplus DATA "file", "format", "type", "nobservations" options. runmplus takes care of all of this for you.
NGroups - number of groups
Variable Options not needed - Generally, the Mplus "names", "useobservations", "usevariables" will not be needed. Use the if and in statements to accomplish the same thing as "Useobservations", and you must specify a variable list so "Usevariables" is assumed to be all of the variables.
categorical - names of categorical dependent variables. Use a string Mplus will understand.
Grouping - name and label values of grouping variable. Put text here just like you would after the GROUPING option in Mplus. For example grouping(gender(1=male 2=female)). In the model statement, refer to these groups as you would in Mplus (eg., model(f by y1 y2; f on x; model male:; f on x; model female:; f on x;)).
Note the semi-colon after model <group>: This helps runmplus parse the command line.
It is possible you'll get an error message or the grouping statement won't be parsed correctly, particularly if you have a long grouping statement. If this happens, try shortening the grouping statement by using shorter group lablels.
DEFINE- The whole point of this program is to allow you to do this kind of stuff in Stata. But, if you want it, you can specify a Mplus option string within the DEFINE option and have Mplus make new variables. This might be useful if you wanted a new variable in the Mplus analysis but not as a part of the master data set.
type - refers to "type" option in Mplus ANALYSIS command. The default is general. Other options include meanstructure, missing, H1, Mcohort, Mixture, Complex, Twolevel, EFA # #, Logistic.
estimator - Mplus will recognize and implement ML, MLM, MLMV, GLS, WLS, WLSM, WLSMV, MLR, MLF, MLM, MLMV, ULS estimators if the model type can support it. If nothing is specified, the Mplus default estimators are used.
matrix - matrix to be evaluated. You can put anything you want but Mplus will only allow "covariance" or "correlation"
tech(number list) - list in number list the technical output to be displayed (just list the numbers)
1 - parameter specifications and starting values for all free parameters.
2 - parameter derivatives.
3 - estimated covariance and correlation matrices for the parameter estimates.
4 - means and covariace and correlation matrices for the latent variables in the model.
5 - optimization history.
6 - optimization in estimating the sample statistics.
7 - sample statistics for each class (when type(mixture) is used) using raw data weighted by the estimated class probabilities.
8 - optimization history when type(mixture) is used).
9 - available when MONTECARLO command is used when type(mixture) is also used, and produces error messages regarding convergence for each replication of the Monte Carlo study.
10 - used in conjunction with TYPE=MIXTURE to request univariate and bivariate model fit information for the categorical dependent variables in the model.
11 - used in conjunction with TYPE=MIXTURE request the Lo-Mendell-Rubin likelihood ratio test of model fit.
12 - used in conjunction with TYPE=MIXTURE to request residuals for observed versus model estimated means, variances, covariances, univariate skewness, and univariate kurtosis.
13 - used in conjunction with TYPE=MIXTURE to request two-sided tests of model fit for univariate, bivariate, and multivariate skew and kurtosis (Mardia’s measure of multivariate kurtosis).
14 - used in conjunction with TYPE=MIXTURE to request a parametric bootstrapped likelihood ratio test.
Author Richard N Jones, ScD Institute for Aging Research Hebrew SeniorLife, Boston, MA jones@mail.hrca.harvard.edu
updates and other STATA+Mplus tools at http://sites.google.com/site/lvmworkshop > /home/runmplus-stuff
Also see
On-line: help for stata2mplus runmplus runmplus_fits read_parameterestimates_general read_parameterestimates read_convergence lli