Log Linear Analysis Flashcards
Chi Square Test of Independence
Categorical IV and DV. Used to compare two proportions (think t-test).
Log Linear Analysis
Used to compare proportions when there are more than two categorical IVs (think factorial ANOVA). Analysis begins with most complex model and deletes significant effects one by one, starting with the highest order interaction. If removing the effect significantly worsens fit, the effect is retained. If non-significant, the effect is deleted.
Assumptions of Log Linear Analysis
Independent observations, linearity, homogeneity of variance, normality, expected frequencies (greater than 5).
Log Linear Analysis and SPSS
K-way effects says whether there were any significant main effects or interactions
Convergence Information says which effects are included in the final model.
Interactions in Log linear Analysis
Split data file and run a chi square analysis for each split. When writing up, discuss results in terms of how data was split.
Odds Ratio
Provides and estimate of effect size for log linear analysis and logistic regression. Calculate the odds for each split of the IV, then work out the ratio. If more than 1, the odds of that outcome increase, if less than 1, the odds of that outcome decrease.