Requesting Output

Normally, Mplus will display some statistics for model fit, parameter estimates and standard errors. Some additional information may be helpful and can be requested with command OUTPUT. I will introduce each keyword in turn, but of course several keywords can follow the OUTPUT command to request all the output desired. As usual, keywords are separated by semicolons.

Sample Statistics

OUTPUT:
SAMPSTAT;

will display sample means, variances, covariances and correlations for continous variables. For other variables, other statistics will be displayed.

Standardization

OUTPUT:
STANDARDIZED;

will display three different standardized solutions: In the first, called STDYX, all latent variables as well as all manifest covariates and outcome variables will be standardized. Other options for standardization are STDY and STD. STDY will standardize all continuous latent variables and all outcome variables; according to the User's Guide (p. 577), this should be used for binary covariates for which standardization is meaningless. STD uses only the variances of the latent continous variables for standardization. If you need only one or two of these solutions, you may use STDYX, STDY or STD instead of STANDARDIZED.

Confidence Intervals

OUTPUT:
CINTERVAL;

will display 1 and 5 per cent confidence intervals. You may also request confidence intervals based on bootstrapping with keywords CINTERVAL(BOOTSTRAP) or CINTERVAL(BCBOOTSTRAP). The latter stands for bias corrected bootstrapping according to MacKinnon, D. P., Lockwood, C. M., & Williams, J. (2004). Confidence Limits for the Indirect Effect. Multivariate Behavioral Research, 39(1), 99-128.

To obtain the latter confidence intervals, you first have to request bootstrapping with the ANALYSIS command.

Modification Indices

To obtain modification indices, use the keyword MODINDICES with the OUTPUT command. By default, Mplus displays only those modification indices with a value of 10 or more. To change this limit, indicate the limit in parentheses.

OUTPUT:
MOD(0);

will display all modification indices (unless you have a just-identified model, in which the MIs will be zero). The modification indices (i.e. change of chi-square of your model if you would set free a parameter) will be displayed together with an estimated parameter change (E.P.C.) and two versions of standardized estimated parameter changes (Std E.P.C. and StdXY E.P.C.).

Factor validity (determinacy)

In factor analysis, one wishes to know how well factor scores represent the true factor. This is often called factor validity; the Mplus name for it is factor score determinacy and is described as the correlation between the estimated and true factor scores. It is obtained with

OUTPUT:
FSDETERMINACY;

Regrettably, factor determinacy can be computed only if all dependent variables (among which are counted all indicators) are continuous.