METHOD Subcommand (MANOVA: Univariate command)

METHOD controls the computational aspects of the MANOVA analysis. You can specify one of two different methods for partitioning the sums of squares. The default is UNIQUE.

UNIQUE. Regression approach. Each term is corrected for every other term in the model. With this approach, sums of squares for various components of the model do not add up to the total sum of squares unless the design is balanced. This is the default if the METHOD subcommand is omitted or if neither of the two keywords is specified.

SEQUENTIAL. Hierarchical decomposition of the sums of squares. Each term is adjusted only for the terms that precede it on the DESIGN subcommand. This decomposition is an orthogonal decomposition, and the sums of squares in the model add up to the total sum of squares.

You can control how parameters are to be estimated by specifying one of the following two keywords that are available on MANOVA. The default is QR.

QR. Use modified Givens rotations. QR bypasses the normal equations and the inaccuracies that can result from creating the cross-products matrix, and it generally results in extremely accurate parameter estimates. This setting is the default if the METHOD subcommand is omitted or if neither of the two keywords is specified.

CHOLESKY. Use Cholesky decomposition of the cross-products matrix. This method is useful for large data sets with covariates entered on the DESIGN subcommand.

You can also control whether a constant term is included in all models. Two keywords are available on METHOD. The default is CONSTANT.

CONSTANT. All models include a constant (grand mean) term, even if none is explicitly specified on the DESIGN subcommand. This setting is the default if neither of the two keywords is specified.

NOCONSTANT. Exclude constant terms from models that do not include the keyword CONSTANT on the DESIGN subcommand.

Example

MANOVA DEP BY A B C (1,4)
  /METHOD=NOCONSTANT 
  /DESIGN=A, B, C
  /METHOD=CONSTANT SEQUENTIAL
  /DESIGN.
  • For the first design, a main-effects model, the METHOD subcommand requests the model to be fitted with no constant.
  • The second design requests a full factorial model to be fitted with a constant and with a sequential decomposition of sums of squares.