Statistical Sciences 2035 Quiz: Advantage of Univariate mixed model over multivariate test
16 Oct 2018
School
Department
Professor
Document Summary
Advantage of univariate mixed model over multivariate test: denominator in univariate mixed model is always larger than in multivariate model easier to reject h0. This advantage is somewhat lost after epsilon correction because it strongly reduces df of denominator, but it still remains larger than in multivariate. Numerator df always stay equal in both models. Less often used test: 1-way ws anova (glm univariate) that leads to the bs table, because it doesn"t recognize it"s a ws analysis always assumes sphericity which is often false. 2-way = 2 within subjects factors (fixed factors) In our example: rest (yes/no) & bonus (yes/no) 4 conditions. To account for correlatedness of observations, we can do a mixed model that incorporates the subjects (pp) as a random factor. Always orthogonal if there is no dropout, because the participants per condition are literally identical = so same amount: therefore: experimental factors are statistically independent & their effects can be estimated independently.
