PB HLTH 245 Lecture Notes - Lecture 5: Scree, Ibm Aix, Dimensionality Reduction
Document Summary
Goal : to find a number of linear combination of the vector of variables x - C xi , xz , xp it to best summarize those variables. Atx = arx , tazxzt - . ay xp = pc. At it is a-- cal , az - - Pc direction out x. aix , aix , . adtx first d 7c. & optimization mg,x varcatx ) maazx var cast x) apai - To preserve as much variability in the variables as possible (first direction , second direction) D var ( atx) = at var cx) a. Cov (atx , b. tx) = atuarcx) b important thing to solution compute { = (cid:8869)vcx ) spectral decomposition of e. { = va vt pc : vj" x. * use scree plot to exam ( find biggest slope drop ) , j - var ( vix ) = rj. Aways center the data for a given data matrix.