logic - Avoiding a for loop in R in an attempt to evaluate percentage of true positives/negatives when using logistic regression -

what got: matrix got predicted probability of outcome (from logistic regression model) , known outcome. curious got 2 regression models , independent test dataset wish compare these 2 models doing this.

> head(matrixcomb)       probcomb outcomb [1,] 0.9999902       1 [2,] 0.9921736       0 [3,] 0.9901175       1 [4,] 0.9815581       0 [5,] 0.7692992       0 [6,] 0.7369990       0 

what want: graph can plot how prediction model yields correct outcomes (one line positives , 1 line negatives) function of cut off value probability. problem unable figure out how without switching perl , use for-loop iterate through matrix.

in perl start @ probability 0.1 , in reach run of for-loop increase value 0.1. in first iteration count probabilities <0.1 , outcome = 0 true negatives, probability < 0.1 , outcome 1 false negatives probability > 0.1 , outcome = 0 false positives , probability > 0.1 , outcome = 1 true positives.

the process repeated , results of each iteration printed [probability, true positives/total positives, true negatives/total negatives]. make easy me print out in open office calc.

the reason asking operation complex me find similar case here on stackoverflow or in tutorial. learn way in efficient manner in r environment.

you can r draw curves based on roc analysis. crude version using rocr package , made prettier

ss <- 1000   # sample size mydf <- data.frame(probcomb = runif(ss)) # predictions illustration mydf$outcomb <- 0 + (runif(ss) < mydf$probcomb) # actuals illustration  library(rocr) pred <- prediction(mydf$probcomb, mydf$outcomb) perfp <- performance(pred, "tpr") perfn <- performance(pred, "tnr") plot(perfp, col="green", ylab="true positive (green) , true negative (red) rates") plot(perfn, col="red", ylab="true negative rate", add=true) 

to produce

if must, can find data in perfp , perfn.