How to create a loop that will add new variables to a pre define glm model

Question

I would like to create a procedure that will add per each loop a new variable (from a pool of variables) to a glm model that allready contains few of the variables that need to be part of the final model.I than would like to have the results of the loop in a list that will contain the glm formula and results.I know how to do it manually (code is written below) but I would be happy to know how to do it automaticaly. Here is a toy dataset and the relevant code to do the task manually:

dat <- read.table(text = "target birds    wolfs     Country
                            0       21         7     a
                            0        8         4     b
                            1        2         5     c
                            1        2         4     a
                            0        8         3     a
                            1        1         12    a
                            1        7         10    b
                            1        1         9  c",header = TRUE)
#birds is a mandatory variable so I'll need to add one of the other   variables in addition to birds
 glm<-glm(target~birds,data=dat)
 dat$glm_predict_response <- ifelse(predict(glm,newdata=dat,   type="response")>.5, 1, 0)
xtabs(~target + glm_predict_response, data = dat)
      glm_predict_response
target 0 1
     0 1 2
     1 0 5
    glm_predict_response
prop.table(xtabs(~target + glm_predict_response, data = dat), 2)
    target         0         1
         0 1.0000000 0.2857143
         1 0.0000000 0.7142857

#manually I would add the next variable (wolfs) to the model and look at the results:
 glm<-glm(target~birds+wolfs,data=dat)
 dat$glm_predict_response <- ifelse(predict(glm,newdata=dat, type="response")>.5, 1, 0)
 xtabs(~target + glm_predict_response, data = dat)
      glm_predict_response
target 0 1
     0 3 0
     1 0 5
 prop.table(xtabs(~target + glm_predict_response, data = dat), 2)
      glm_predict_response
target 0 1
     0 1 0
     1 0 1

In the next loop I would add the variable "country" and do the same procedure, In the real life I have hundreds of variables so turning it to an automatic proccess would be great.

Answer 1

You can try something like

    list_1=list(NA)
    list_2=list(NA)
    for (i in 2 :ncol(dat)){
      dat1=dat[,1:i]
      glm<-glm(target~.,data=dat1)
      dat1$glm_predict_response <- ifelse(predict(glm,newdata=dat1,   type="response")>.5, 1, 0)

      list_1[[i-1]]=xtabs(~target + glm_predict_response, data = dat1)
      names(list_1)[i-1]=do.call(paste,as.list(colnames(dat1)[c(-1,-ncol(dat1))]))

      list_2[[i-1]]=prop.table(xtabs(~target + glm_predict_response, data = dat1), 2)
      names(list_2)[i-1]=do.call(paste,as.list(colnames(dat1)[c(-1,-ncol(dat1))]))}

But you need to have col in right order.

Answer 2

I would do it using update to update the formula each time in the loop:

#initiate formula
myform <- target~1
for ( i in c('birds', 'wolfs' , 'Country')) { 
    #update formula each time in the loop with the above variables
    #this line below is practically the only thing I changed
    myform <- update(myform,  as.formula(paste('~ . +', i)))
    glm<-glm(myform,data=dat)
    dat$glm_predict_response <- ifelse(predict(glm,newdata=dat,   type="response")>.5, 1, 0)
    print(myform)
    print(xtabs(~ target + glm_predict_response, data = dat))
    print(prop.table(xtabs(~target + glm_predict_response, data = dat), 2))

}

Output:

target ~ birds
      glm_predict_response
target 0 1
     0 1 2
     1 0 5
      glm_predict_response
target         0         1
     0 1.0000000 0.2857143
     1 0.0000000 0.7142857

target ~ birds + wolfs
      glm_predict_response
target 0 1
     0 3 0
     1 0 5
      glm_predict_response
target 0 1
     0 1 0
     1 0 1

target ~ birds + wolfs + Country
      glm_predict_response
target 0 1
     0 3 0
     1 0 5
      glm_predict_response
target 0 1
     0 1 0
     1 0 1