I am doing Principal Component Analysis (PCA) on assets data for household
income prediction. The problem is that the assets data are rank ordered
(usually binary ... possess car/don't possess car), so the normal correlation
is inappropriate for the calculation of the PCA. Instead one has to use the
polychoric correlation coefficient. It uses the "random.polychor.pa" package.
Scenario 1
If i only use PCA without using polychoric correlation
assets.pca <- princomp(covmat = assets, scores=T)
assets$income<-predict(assets2.pca)[,1]
# these predict the coefficient for income for each observation
Scenario 2
but when i run "hetcor" (for polychoric correlation coefficient) , and
then princomp (for PCA) and finally predict ( to predict
the coefficient for income for each observation).
hetcor and princomp runs fine but predict command doesn't work and
it doesn't predict value for each observation
library(epicalc)library(foreign)library(polycor)
assets.mat <- hetcor(assets, use="complete.obs")[[1]] assets.pca <-
princomp(covmat = assets.mat, scores=T)assets$income<-predict(assets.pca)[,1]
the problems are why this "pridict" commend throwing error in scenario 2
and not in scenario 1?Is there any other way to predict the values for each
observation?
thanks a lotmm
[[alternative HTML version deleted]]
______________________________________________
R-help@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.
