PHIL 1 Lecture Notes - Dependent And Independent Variables, Sphericity, Trend Analysis
Document Summary
Local sphericity: when not all, but two variances of differences are similar: this assumption only exists in univariate model, not in multivariate model. Sums of squares, mean squares, f-ratio: ss(total)=ss(between)+ss(within, ss(within)=ss(model)+ss(residual, f=ms(model)/ms(residual) Univariate test (spss: tests of within-subjects effects: we need to enter data as univariate data format, multiple records per subject. Include participant (pp) as random factor necessary, otherwise error variance will be overestimated because spss acts as if all observations come from independent data points: variables: pp, condition, y. If we have three conditions, there will be three rows per subjects. Score: do familiar anova and get test of between-subjects effects table, h0: all population means are equal, should h0 be rejected and factor has more than 2 levels, do pairwise comparisons with bonferroni. Multivariate tests (spss: multivariate tests: we need to enter data as multivariate data format, each participant has only one row, for each level of dependent variable, we create an own variable.
