TEST Subcommand (GLM: Univariate command)

The TEST subcommand allows you to test a hypothesis term against a specified error term.

• TEST is valid only for univariate analyses. Multiple TEST subcommands are allowed, with each subcommand being executed independently.
• You must specify both the hypothesis term and the error term. There is no default.
• The hypothesis term is specified before the keyword VS and must be a valid effect that is specified or implied on the DESIGN subcommand.
• The error term is specified after the keyword VS. You can specify either a linear combination or a value. The linear combination of effects takes the general form: coefficient*effect +/– coefficient*effect ...
• All effects in the linear combination must be specified or implied on the DESIGN subcommand. Effects that are specified or implied on DESIGN but not listed after VS are assumed to have a coefficient of 0.
• Duplicate effects are allowed. GLM adds coefficients associated with the same effect before performing the test. For example, the linear combination 5*A–0.9*B–A is combined to 4*A–0.9B.
• A coefficient can be specified as a fraction with a positive denominator (for example, 1/3 or –1/3 are valid, but 1/–3 is invalid).
• If you specify a value for the error term, you must specify the degrees of freedom after the keyword DF. The degrees of freedom must be a positive real number. DF and the degrees of freedom are optional for a linear combination.

Example

GLM DEP BY A B
  /TEST = A VS B + A*B
  /DESIGN = A, B, A*B.

• A is tested against the pooled effect of B + A*B.