国外心理学-设计和分析课件empty【荐】.pdfVIP

Unequal Cell Sizes Psychology 3256 Introduction • Most designs assume equal ns • except for the simplest ones, like say one way ANOVA • Well let’s say you have unequal ns, what then? What to do.. • Perhaps estimate the missing value(s) • use the mean or media maybe • do you include interactions or not? • No matter what, you have lost df There are other ways • Most of our analyses use Type I SS • They are fine if ns are equal • bad when unequal • Type II SS • They are good when there is a missing value, but…. • but bad when there is an interaction Type III SS • You may not know if there is an interaction or not • so if you use Type III you are safe • while we have learned Type I, analysis software uses Type III • If ns are equal Type I=Type II=Type III Type IV • Well, they can work with empty cells • The problem with empty cells is umm, there is no unique solution Empty Cells A1 A2 So, what goes there? B1 4 2 B2 2