Overview (ANOVA command)
ANOVA performs analysis of variance for factorial designs. The default is the full factorial model if there are five or fewer factors. Analysis of variance tests the hypothesis that the group means of the dependent variable are equal. The dependent variable is interval-level, and one or more categorical variables define the groups. These categorical variables are termed factors. ANOVA also allows you to include continuous explanatory variables, termed covariates. Other procedures that perform analysis of variance are ONEWAY, SUMMARIZE, and GLM. To perform a comparison of two means, use TTEST.
Options
Specifying Covariates. You can introduce covariates into the model using the WITH keyword on the VARIABLES subcommand.
Order of Entry of Covariates. By default, covariates are processed before main effects for factors. You can process covariates with or after main effects for factors using the COVARIATES subcommand.
Suppressing Interaction Effects. You can suppress the effects of various orders of interaction using the MAXORDERS subcommand.
Methods for Decomposing Sums of Squares. By default, the regression approach (keyword UNIQUE) is used. You can request the classic experimental or hierarchical approach using the METHOD subcommand.
Statistical Display. Using the STATISTICS subcommand, you can request means and counts for each dependent variable for groups defined by each factor and each combination of factors up to the fifth level. You also can request unstandardized regression coefficients for covariates and multiple classification analysis (MCA) results, which include the MCA table, the Factor Summary table, and the Model Goodness of Fit table. The MCA table shows treatment effects as deviations from the grand mean and includes a listing of unadjusted category effects for each factor, category effects adjusted for other factors, and category effects adjusted for all factors and covariates. The Factor Summary table displays eta and beta values. The Goodness of Fit table shows R and R 2 for each model.
Basic Specification
- The basic specification is a single VARIABLES subcommand with an analysis list. The minimum analysis list specifies a list of dependent variables, the keyword BY, a list of factor variables, and the minimum and maximum integer values of the factors in parentheses.
- By default, the model includes all interaction terms up to five-way interactions. The sums of squares are decomposed using the regression approach, in which all effects are assessed simultaneously, with each effect adjusted for all other effects in the model. A case that has a missing value for any variable in an analysis list is omitted from the analysis.
Subcommand Order
- The subcommands can be named in any order.
Operations
A separate analysis of variance is performed for each dependent variable in an analysis list, using the same factors and covariates.
Limitations
- A maximum of 5 analysis lists.
- A maximum of 5 dependent variables per analysis list.
- A maximum of 10 factor variables per analysis list.
- A maximum of 10 covariates per analysis list.
- A maximum of 5 interaction levels.
- A maximum of 25 value labels per variable displayed in the MCA table.
- The combined number of categories for all factors in an analysis list plus the number of covariates must be less than the sample size.