variable selection using residual difference
Suggestion: Don't do this! I suggest that you consult with a local statistician or post to a statistical website like stats.stackexchange.com for what might be sensible procedures for variable selection (a complex and controversial topic!) and why what you propose is or is not a good idea (don't trust me!). Cheers, Bert Bert Gunter "The trouble with having an open mind is that people keep coming along and sticking things into it." -- Opus (aka Berkeley Breathed in his "Bloom County" comic strip ) On Fri, Mar 18, 2016 at 9:00 AM, Hassan, Nazatulshima
<Nazatulshima.Hassan at liverpool.ac.uk> wrote:
I have the following example dataset
set.seed(2001)
n <- 100
Y <- c(1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0)
X1 <- sample(x=c(0,1,2), size=n, replace=TRUE, prob=c(0.1,0.4,0.5))
X2 <- sample(x=c(0,1,2), size=n, replace=TRUE, prob=c(0.5,0.25,0.25))
X3 <- c(0,2,2,2,2,2,2,2,0,2,0,2,2,0,0,0,0,0,2,0,0,2,2,0,0,2,2,2,0,2,0,2,0,2,1,2,1,1,1,1,1,1,1,1,1,1,1,0,1,2,2,2,2,2,2,2,2,2,2,2,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,1,1,1,1,1,1,1,1,1,0,0,0,0,0,0,0,1,0,0,0,0)
dat <- data.frame(Y,X1,X2,X3)
I fit a logistic regression model to each of the variable to rank them based on the residual difference (highest to lowest). To simplify I got the rank as X3,X1 and X2. Then, I fit a second order model as follows and again calculate the res_dif :
mod1 <- glm(Y~X3+X1, family=binomial, data=dat)
mod1$null.deviance-mod1$deviance
mod2 <- glm(Y~X3+X2, family=binomial,data=dat)
mod2$null.deviance-mod2$deviance
Again, I will rank the model based on res_dif (highest to lowest). So here, I choose mod2. From there I will fit the third order model as follows :
mod3 <- glm(Y~X3+X2+X1, family=binomial, data=dat)
mod3$null.deviance-mod3$deviance
Basically, this continues until it fits the maximum number of variables that you have in the data.
My aim is to do variable selection based on res_dif instead of AIC, BIC or R2. Since my actual dataset is dealing with 100 of variables, I wonder how can I apply this using loop function.
Any suggestions would be appreciated.
Kind Regards
Shima
[[alternative HTML version deleted]]
______________________________________________ R-help at r-project.org mailing list -- To UNSUBSCRIBE and more, see 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.