Statistical Sciences 2035 Quiz: Test main effects
16 Oct 2018
School
Department
Professor
Document Summary
Test main effects (instead of simple effects): comparing marginal means. Relative frequencies of one variable differ between the levels of the other variable. There is a statistical relationship between both factors: both factors are correlated, the various effects we wish to. Confounding problem is demonstrated when comparing 1-way (without sex) and 2-way analysis: Because 1-way anova includes no correction for the confounding effect of gender. Ex: 1-way: mean in divorced condition is lower than in married condition; but this difference exists, because men on average score lower. However, in the descriptives tale of 1-way analysis you don"t see the men/women ratio, so this confounding variable is ignored. Once overlap with gender is filtered out, the differences between status means are no longer significant. Can be done in a variety of ways; conventional approach: Focusing only on the aspect of variation that does not overlap with the other effects: will eliminate the confounder. Venn diagrams do not mathematically represent situation correctly.
