Example of an HMLM model

To create a new command file based on nys.mdm, select Create a new model using an existing MDM file option from the File menu.

Select the previously created MDM file from the \examples\chapter9 folder when prompted to do so and click OK to return to the main WHLM window.

The level-1 variables for which information is available in the MDM file are displayed on the left of the main window. As a first step, select the variable ATTIT as the outcome variable by clicking on the variable name and selecting outcome variable on the pop-up menu as shown below.

The unconditional models with both unrestricted and homogeneous level-1 variance structures are now displayed by default in the main window. Add AGE11 as an uncentered predictor at level-1, as shown below.

Note that the model now displayed is still unconditional at level-2.  To add the covariate FEMALE to both the intercept and slope equations, click the Level-2 button on the left of the window to display the level-2 variables available. Next, click on the intercept equation, add FEMALE uncentered to the intercept equation, as shown below.

To add the FEMALE covariate to the slope equation too, first click on this equation and then add FEMALE in the same way as was done with the intercept equation. Also activate the error term for the slope equation by clicking on it. The complete model is now displayed in the main window, as shown below.

In order to run a latent variable analysis with HMLM for the 2-level model, the latent variable must be specified. Select the Estimation Settings option from the Other Settings menu to open the Estimation Settings – HMLM dialog box, then access the Latent Variable Regression dialog box by clicking on the Latent Variable Regression button provided in the dialog box shown below.

On the Latent Variable Regression dialog box, click the radio buttons to indicate that AGE11 is the outcome, and the intercept is the predictor for this example.

, the level of tolerance at age 11, is used to model the linear growth rate, represented by . Both and  are latent variables, as they are free of measurement error. The effect of gender on the linear growth rate is assessed, after controlling for the initial status at age 11.

Note that the homogeneous level-1 variance option (see the Basic Model Specifications – HMLM dialog box) is used for this model. This means that the results obtained for this example will be identical to those obtained using HLM2 instead of HMLM. Also, results for both the unrestricted model and the model with homogeneous level-1 variance  will be produced by HMLM.

Finally, save the newly created command file by selecting Save As on the File menu and assigning a name, in this case nys2.mlm, to the command file. Click Save to return to the main WHLM window and click Run Analysis to start the analysis.