ERROR Subcommand (MANOVA: Univariate command)

ERROR allows you to specify or change the error term that is used to test all effects for which you do not explicitly specify an error term on the DESIGN subcommand. ERROR affects all terms in all subsequent designs, except terms for which you explicitly provide an error term.

WITHIN. Terms in the model are tested against the within-cell sum of squares. This specification can be abbreviated to W. This setting is the default unless there is no variance within cells or a continuous variable is named on the DESIGN subcommand.

RESIDUAL. Terms in the model are tested against the residual sum of squares. This specification can be abbreviated to R. This specification includes all terms not named on the DESIGN subcommand.

WITHIN+RESIDUAL. Terms are tested against the pooled within-cells and residual sum of squares. This specification can be abbreviated to WR or RW. This setting is the default for designs in which a continuous variable appears on the DESIGN subcommand.

error number. Terms are tested against a numbered error term. The error term must be defined on each DESIGN subcommand. For a discussion of error terms, see DESIGN Keyword (MANOVA: Univariate command) .

  • If you specify ERROR=WITHIN+RESIDUAL and one of the components does not exist, MANOVA uses the other component alone.
  • If you specify your own error term by number and a design does not have an error term with the specified number, MANOVA does not carry out significance tests. MANOVA will, however, display hypothesis sums of squares and, if requested, parameter estimates.

Example

MANOVA DEP BY A(1,2) B(1,4)
  /ERROR = 1
  /DESIGN = A, B, A BY B = 1 VS WITHIN
  /DESIGN = A, B.
  • ERROR defines error term 1 as the default error term.
  • In the first design, A by B is defined as error term 1 and is therefore used to test the A and B effects. The A by B effect itself is explicitly tested against the within-cells error.
  • In the second design, no term is defined as error term 1, so no significance tests are carried out. Hypothesis sums of squares are displayed for A and B.