PSYC3301 Chapter Notes - Chapter 7-8: F-Test, Sphericity, Repeated Measures Design
Theory of anova
- comparing the amount of variability explained by the model (mean differences) to
the error in the model (individual differences): this is called the F ratio (variability
between means and individual observations)
- within groups (bad shit): variance associated with the effect of interest
- between groups (good shit): variance associated with the effect of interest + other
sources we dk about
- cohen’s d effect size .1 small .3 medium .5 large
- coefficient of determination : proportion of variance in the dependent acc for by the
independent
Between Subjects ANOVA
- Assumption: Test of homogeneity
Less than 3 means
- equal sample size, homogeneity is satisfied: Fisher’s LSD for multiple comparison
More than 3 means
- homogeneity is satisfied: Turkey Kramer (bit conservative) or Fisher Hayter (SPSS
doesnt do)
- if violates homogeneity and unequal sample sizes: Welch’s F test / Brown - Forsythe
F test (ANOVA)
- multiple comparisons: Games-Howell procedure
Within Subjects ANOVA (repeated measures)
- extension of the paired sample t test but anova is more than 3 means
- Assumption: Mauchly’s Test of Sphericity (p must be >.5) epsilon value of .00 - 1.0
(complete sphericity)
- partial n2: proportion of variance in the dependent variable accounted for by the
independent variable excluding the variance accounted for by the correlations
between conditions
Less than 3 means
- multiple comparison: Fisher’s protected LSD
More than 3 means
- multiple comparisons: Bonferroni (options not post hoc)
Violated test of sphericity
- look at the Huynh-Feldt results
find more resources at oneclass.com
find more resources at oneclass.com
Document Summary
Comparing the amount of variability explained by the model (mean differences) to the error in the model (individual differences): this is called the f ratio (variability between means and individual observations) Within groups (bad shit): variance associated with the effect of interest. Between groups (good shit): variance associated with the effect of interest + other sources we dk about. Cohen"s d effect size . 1 small . 3 medium . 5 large. Coefficient of determination : proportion of variance in the dependent acc for by the independent. Equal sample size, homogeneity is satisfied: fisher"s lsd for multiple comparison. Homogeneity is satisfied: turkey kramer (bit conservative) or fisher hayter (spss doesnt do) If violates homogeneity and unequal sample sizes: welch"s f test / brown - forsythe. Extension of the paired sample t test but anova is more than 3 means. Assumption: mauchly"s test of sphericity (p must be >. 5) epsilon value of . 00 - 1. 0 (complete sphericity)
