st: RE: weighted Factor Analysis

["Verkuilen, Jay" <JVerkuilen@gc.cuny.edu>]

Rafaelcla wrote:

>>>I have a dataset of food acquisition data (indicating the average per capita acquisition of some food groups) where the unit of analysis are clusters of household (basically it is an aggregated dataset of household food acquisition data). The original dataset contain a "probability weight variable" (PW) which has been used for all the analysis so far. The problem is that I can not use the "pw" to perform Factor Analysis (starting with fac command). Should I just not weight the data while performing factor analysis?<<<

Stata's factor analysis program analyzes correlation matrices. You could use -factormat- to do the calculation on a computed correlation matrix that used weights. -cor- allows frequency weights and analytic weights. You may need to do some tricks to get the right value. 

In general you should weight---not doing so probably won't give you what you want because it won't be applicable to the population at hand. 

Jay

<<winmail.dat>>