Overview (MANOVA: Univariate command)

This section describes the use of MANOVA for univariate analyses. However, the subcommands that are described here can be used in any type of analysis with MANOVA. For additional subcommands that are used in multivariate analysis, see MANOVA: Multivariate. For additional subcommands that are used in repeated measures analysis, see MANOVA: Repeated Measures. For basic specification, syntax rules, and limitations of the MANOVA procedures, see MANOVA.

Options

Design Specification. You can use the DESIGN subcommand to specify which terms to include in the design. This ability allows you to estimate a model other than the default full factorial model, incorporate factor-by-covariate interactions, indicate nesting of effects, and indicate specific error terms for each effect in mixed models. You can specify a different continuous variable as a dependent variable or work with a subset of the continuous variables with the ANALYSIS subcommand.

Contrast Types. You can specify contrasts other than the default deviation contrasts on the CONTRAST subcommand. You can also subdivide the degrees of freedom associated with a factor (using the PARTITION subcommand) and test the significance of a specific contrast or group of contrasts.

Optional Output. You can choose from a variety of optional output on the PRINT subcommand or suppress output using the NOPRINT subcommand. Output that is appropriate to univariate designs includes cell means, design or other matrices, parameter estimates, and tests for homogeneity of variance across cells. Using the OMEANS, PMEANS, RESIDUAL, and PLOT subcommands, you can also request tables of observed and/or predicted means, casewise values and residuals for your model, and various plots that are useful in checking assumptions. In addition, you can use the POWER subcommand to request observed power values (based on fixed-effect assumptions), and you can use the CINTERVAL subcommand to request simultaneous confidence intervals for each parameter estimate and regression coefficient.

Matrix Materials. You can write matrices of intermediate results to a matrix data file, and you can read such matrices in performing further analyses by using the MATRIX subcommand.

Basic Specification

  • The basic specification is a variable list that identifies the dependent variable, the factors (if any), and the covariates (if any).
  • By default, MANOVA uses a full factorial model, which includes all main effects and all possible interactions among factors. Estimation is performed by using the cell-means model and UNIQUE (regression-type) sums of squares, adjusting each effect for all other effects in the model. Parameters are estimated by using DEVIATION contrasts to determine whether their categories differ significantly from the mean.

Subcommand Order

  • The variable list must be specified first.
  • Subcommands that are applicable to a specific design must be specified before that DESIGN subcommand. Otherwise, subcommands can be used in any order.

Syntax Rules

  • For many analyses, the MANOVA variable list and the DESIGN subcommand are the only specifications that are needed. If a full factorial design is desired, DESIGN can be omitted.
  • All other subcommands apply only to designs that follow. If you do not enter a DESIGN subcommand, or if the last subcommand is not DESIGN, MANOVA uses a full factorial model.
  • Unless replaced, MANOVA subcommands (other than DESIGN) remain in effect for all subsequent models.
  • MISSING can be specified only once.
  • The following words are reserved as keywords or internal commands in the MANOVA procedure: AGAINST, CONSPLUS, CONSTANT, CONTIN, MUPLUS, MWITHIN, POOL, R, RESIDUAL, RW, VERSUS, VS, W, WITHIN, and WR. Variable names that duplicate these words should be changed before you invoke MANOVA.
  • If you enter one of the multivariate specifications in a univariate analysis, MANOVA will ignore it.

Limitations

  • A maximum of 20 factors is in place.
  • A maximum of 200 dependent variables is in place.
  • Memory requirements depend primarily on the number of cells in the design. For the default full factorial model, the number of cells equals the product of the number of levels or categories in each factor.