Special Designs in Manova

1
Special Designs in Manova

• Contrasts
• Simple Effects
• Doubly Multivariate Designs

2
Contrasts

• Between-subjects contrasts can be
• Univariate (if single DV or two-level w/s factor)
• Both SINGLEDF and PARAM(ESTIM) show them
• Multivariate (if 2 DVs or more-level w/s
  factor)
• Only SINGLEDF shows the discriminant functions
• Within-subject contrasts are either univariate
  differences (k 2) or single discriminant
  functions of multivariate contrasts analyzed
  simultaneously (e.g., LIN, QUAD).

3
Simple Effects

• When the hypothesis requires it
• When the meaning of a factor discourages
  averaging across levels of that factor
• E.g., no-caffeine control vs. experimental groups
  receiving caffeine random sample of controls vs.
  criminal sociopaths
• What kind of population would that generalize to?
• When n-way interaction is too difficult to
  interpret
• But do not use simple effects as follow-ups to
  interactions use them instead.

4
Interactions
1_23 x LIN, 1_23 x QUAD,2_3 x LIN, 2_3 x QUAD
MANOVA DV by G(1,3) T(1,3)
5
Simple Effects
G1 x LIN, G1 x QUAD,G2 x LIN, G2 x QUAD, G3 x
LIN, G3 x QUAD

6
SPSS Rules

• Simple effects are requested in the /DESIGN (b/s)
  or /WSDESIGN (w/s) command
• Only one type (b/s vs. w/s) can be requested at
  one time, though multiple versions. Effects of
  the other type are crossed with the specified
  simple effects)
• MWITHIN A(k) defines levels within which effects
  of the other type are run
• A WITHIN B(k) defines one factor within levels of
  another
• /DESIGN commands are considered exhaustive no
  other effects are tested. /WSDESIGN tests are
  automatically crossed with any between-subjects
  factor.

L15.html
7
Doubly Multivariate Designs

• Two sources of multivariateness
• W/S factors with 2 levels
• More than one DV (measure) per W/S cell
• Example RTs and errors across 3 trials

TRIAL1 TRIAL2 TRIAL3
L15.html
8
Multiple-Measure W/S Designs

• W/S factors with 2 levels (univariate), but
  multivariateness due to more than one DV
  (measure) per W/S cell
• Example HR and SC within a 2 x 2 design

L15.html
9
Covariates

• MANOVA DV by G(1,3) WITH COV
• /METHOD SEQU
• L12_covariates.html

COMPUTE C2 C MANOVA X1 X2 by G(1,3) WITH C C2
/WSFACT RETEST(2) /METHOD SEQU
10
(No Transcript)