Multivariate GLM is a technique to conduct Analysis of Variance for experiments having more than one dependent variable. This procedure can also be used for multivariate regression analysis with more than one dependent variable. The main dialog box asks for Dependent Variables(responses), Fixed Effect Factors, Random Effect Factors, Covariates (continuous scale), and WLS (Weighted Least Square) weight.
The sub-menus include:
- Model - This is the dialog box for creating your model. The default is a full factorial. You can customize this to only include the interactions that you want. You may want to customize if you want covariate interaction included, as this is not included in the full factorial. Sum of Squares is also set here. If there are no missing cells, Type III is most commonly used. If you have missing cells, you must use Type IV.
- Contrasts - These are used to test for differences among the levels of a factor. They are like Post Hoc, but are specified prior to the experiment. This is only used if you have data with more than two levels.
- Plots - This is chosen if you want a profile (line) plot of the marginal means.
- Post Hoc - Here you can choose a test and use it to determine which means differ. Tukey is generally used if you have a large number of comparisons. For a small number of comparisons, Bonferroni can also be used. This is only used if you have more than two levels for the variable.
- Save- If you want to save any of your output variables (i.e. predicted values, residuals, diagnostics), you must choose this. You can save these to your data editor window or save them to a new file.
- Options- Options let you select the factors for estimates of marginal means. You can also choose additional tests here (such as the Homogeneity test to confirm the assumption of equal variance) and set the significance level.
For the demonstrations purpose we will use the Plastics data set, which has 3 dependent variables, tear resistance, gloss and opacity, and two factors which are extrusion and additive amount.
- To analyze a multivariate GLM, Click Analyze > General Linear Model > Multivariate.
- You should see the dialog box below. Add tear resistance, gloss and opacity into dependent variables textbox and extrusion and additive amount as two factors as per the dialog box below.
- Click Options to change some settings, add additive and extrusion rate into display means for textbox and also choose homogeneity test.
- Once the process is done, click Continue.
- Click OK to see the results as per below: