STA 304 Lecture Notes - Lecture 1: Parametric Statistics, X&Y, Affix
Document Summary
What type - 1: replace each x/y by rank order among the xs/ys. 2: define dk as the dif in rank of kth xy pair. Perfect pos linear association: dk=0vk; no lin assoc: random signs and mag; perf neg: dk large for some and small for others. Ho: =0 ha: 0 (no sig asoc) , >0 (pos asoc bw x/y), <0 (neg asoc bw x/y) : r=1-(2d/dm)=1-[(6 d2)/n(n2-1)] (-1cv, significant: reject ho. For tied ranks: take mean. (note: r is the actual coef of rank) How much power is lost by an analogous nonparametric test being used while the conditions for a parametric test are met. Ape: provides a measure of relative performance w/ large samples used. Kendall"s coef of concordance: used for more than 2 ranks (k>2). : ho=the rankings are unrelated/indep for k rankings of n individuals (there is no association) and ha: association.