PSY3402 Methods and Statistics Content Notes and Example Questions

Week 7: ANOVA with Between Subjects designs

Example/ Group members will be less likely to perceive conflict with another group when
their group has high status
IV: status (high vs low) = two conditions
DV: perceptions of intergroup conflict
Analysis option: t-test OR one-way ANOVA

Example/ Group members will be less likely to perceive conflict with another group when
their group has high status or equal status
IV: status (high vs equal vs low) = three conditions
DV: perceptions of intergroup conflict
Analysis option = one-way ANOVA
- Allows an overall test of the effect of the IV and tests of the specific comparisons
between the individual conditions

Analyse  General Linear Model  Univariate  IV in IV DV in DV  Options: descriptive
statistics, estimates of effect sizes  EMMeans compare main effects  OK

Paste for Syntax
/EMMEANS = TABLES (IngroupStatus) COMPARE ADJ (LSD)
- This command provides a table of means and pairwise comparisons between
conditions
o Pairwise comparisons test the difference between two specific conditions
within an IV (usually that has 3+ conditions); so, comparing conditions 1
and 3 would involve a pairwise comparison
- Tests the effect of one IV at specific levels of another IV/IV’s eg. What is the effect of
X at different levels of Y?
- Simple main effects: univariate box is the data
/CONTRAST = (IngroupStatus) = SPECIAL (2, -1, -1)
- Asks for a specific contrast between low (2) and equal and high combined (-1,-1)
- Tests the difference between conditions/combinations of conditions within a given IV
eg. Is the difference between conditions 1 vs 2 and 3 of variable X significant?
Crucial Output Features
- F value = F ratio = amount of variance you’re looking at
- Sum of squares divided by df gives the Mean square
- Mean square divided by Error = F ratio

More than one IV affects DV and that effect of one IV depends on another IV = interaction
effect
- The effect of one IV is different at different levels of another IV

Example/ Group members will be less likely to perceive conflict with another group when
their group has high status but only when their ingroup has high status, but only when
addressing an outgroup audience, rather than an ingroup audience
IV1: status (high vs low)
IV2: audience (ingroup vs outgroup)

,DV: perceptions of intergroup conflict
Analysis option = two-way ANOVA (will give you 3 effects)
- Tests the main effects of 2 IV’s – effect of status across audience
- Tests interaction b/w 2 IV’s – extent to which effect of status varies depending on
audience

The specific prediction requires analysis of the effect of status at each level of audience
variable, therefore a simple main effect analysis must be done
Simple main effect analysis  tests the main effects of IV1 at each different level of IV2

Syntax
/EMMEANS = TABLES (IngroupStatus*Audiencemap) COMPARE (IngroupStatus) ADJ (LSD)
- This asks for the simple main effect of status at each level of audience

Analyse  General Linear Model  Univariate  Conflict = DV, Status = IV, Audience = IV
 Options: descriptive stats, estimates of effect sizes  EM Means = status, audience and
status*audience, Compare main effects  Paste (for Syntax)  OK

Output:
Tests of between subjects effects:
- Status = main effects are significant
- Audience = main effects are non significant
- Status*Audience = interaction is non significant

F(2,101) = 2.97, p = 0.056, ηρ2 = 0.06

Univariate tests (need to run Syntax to get this):
- Outgroup = simple main effect of status is ONLY significant in outgroup audience

F(2,101) = 4.56, p = 0.013, ηρ2 = 0.08

Estimates are the means in which this is based
- Ingroup audience low status is low 0.3, ingroup audience high status is 0.7 = not
significantly different
- Effect of status in outgroup audience condition is significant M=1.010 low and
M=0.303 high
- Graphically gives a visual representation  when ingroup has low status, perceived
conflict is higher than high status

Pairwise comparisons:
- Outgroup = when status is low, equal and high are significantly different but only in
the outgroup

Equal p = 0.003, High p = 0.048
- Descriptively indicates what we think but the difference b/w ingroup and outgroup
isn’t wholly significant

, Good to plot a bar graph

Example/ Group members will be less likely to perceive conflict with another group when
their group has high status but only when their ingroup has high or equal status, but only
when addressing an outgroup audience, rather than an ingroup audience
IV1: status (high vs equal vs low)
IV2: audience (ingroup vs outgroup)
DV: perceptions of intergroup conflict
Analysis option = two-way ANOVA

Syntax
/EMMEANS = TABLES (IngroupStatus*Audiencemap) COMPARE (IngroupStatus) ADJ (LSD)

Example/ Group members will be less likely to perceive conflict with another group when
their group has high status but only when their ingroup has high status, but only when
addressing an outgroup audience, and when perceived threat is low rather than high
IV1: status (high vs low)
IV2: audience (ingroup vs outgroup)
IV3: threat (high vs low)
- The effect of IV1 will depend on IV2 but only at specific levels of IV3 = three-way
interaction effect
DV: perceptions of intergroup conflict
Analysis option = three-way ANOVA
- Tests main effects of 3 IV’s, 3 two-way interactions b/w each pair of IV’s
(statusXaudience, threatXaudience, statusXthreat)
- Tests a three-way interaction b/w all IV’s

Example/ Analysis of desire for hostile action scores 1) Status would affect desire for hostile
action 2) Only the case when addressing outgroup audience
IV1:status (high vs equal vs low)
IV2: audience (ingroup vs outgroup)
DV: hostile action
Analysis option = 3x2 ANOVA

Analyse  General Linear Model  Univariate  Hostile Action = DV, Status = IV, Audience
= IV  Options: descriptive statistics, estimates of effect sizes  EMMeans: all of them in
and click compare main effects  Paste for Syntax  OK

A 3(status: high vs equal vs low) X 2(audience: ingroup vs outgroup) between participants
ANOVA revealed that the main effects of ingroup status F(2,101) = 0.37, p = 0.694, ηρ2 =
0.01 and audience F(1,101) = 1.83, p = 0.179, ηρ2 = 0.18 were not significant.

The two-way interaction was also not significant F(2,101) = 1.74, p = 0.180, ηρ2 = 0.30.