r - Problems with Gamma Regression Model after doing PCA decomposition -
i unfamiliar pca's , implementing them in r. have data , want apply pca decomposition way deal co-linearity. after applying pca analysis, want use gamma regression model, not sure how fix error keep getting. post code , results below:
library(cosmophotoz) totaldata<-merge(phat0test, phat0train, all=true) summary(totaldata) #gamma regression without pca done on phat0train, included in cosmophotoz gamma1<-glm(redshift~., family=gamma(link="log"), data=phat0train) summary(gamma1) #works fine #new subset of data set.seed(0508) sample<-sample(1:nrow(totaldata), 0.90*nrow(totaldata)) test<-totaldata[sample,] train<-totaldata[-sample,] #pca on subsets mytest<-subset(test, select=-c(redshift)) pcatest <- princomp(mytest,cor=t) summary(pcatest, loadings=t) mytrain<-subset(train, select=-c(redshift)) pcatrain <- princomp(mytrain,cor=t) summary(pcatrain, loadings=t) #formula based on pca loadings keep @ least 99.6% variance form<-redshift~(poly(comp.1, 2))^2 + poly(comp.2, 2)*comp.3*comp.4*comp.5*comp.6 #gamma regression post pca gammafit<-glm(form, family=gamma(link="log"), data=pcatrain)
the error message posted below. of right now, have better understanding on why getting error. reading, has pca's being list , not data fram. still don't know how fix , know if correct way use pca , regression.
error in as.data.frame.default(data) : cannot coerce class ""princomp"" data.frame
Comments
Post a Comment