0

I have been trying to calculate some measures about specificity and sensitivity.
The SDMTools R package worked great to get the overall estimates.

library(SDMTools)

a <- c(1,1,1,1,0,0,0,1,0,1,1,1) #observed
b <- c(1,1,0,0,1,0,0,1,0,0,1,1) #predicted

accuracy(a, b)
#       threshold       AUC omission.rate sensitivity specificity prop.correct     Kappa
#1            0.5    0.6875         0.375       0.625        0.75    0.6666667 0.3333333

Unfortunately, I was not able to get the raw numbers for TRUE POSITIVE,TRUE NEGATIVE, FALSE POSITIVE, FALSE NEGATIVE. I would like to have them in four different columns. Does anyone know a package or a function that can do that? The result I would like to get is below.

#TP  TN  FP  FN
#5    3   1   3
Michael Matta
  • 394
  • 2
  • 16
  • 1
    You can use `table(a,b)`, or `data.frame(table(a,b))` if you prefer a different format. True positives would be cases where a = b = 1, true negatives where a = b = 0, etc. – AntoniosK Oct 09 '18 at 16:36
  • 1
    `caret::confusionMatrix(as.factor(a), as.factor(b))` gives you the all the details from above plus the confusion matrix (table) – phiver Oct 09 '18 at 16:42

1 Answers1

2

You could use a custom function like this one below to calculate the true / false classification counts: 

accuracy_table <- function(obs, pred){
  data.frame(TP = sum(obs == 1 & pred == 1),
             TN = sum(obs == 0 & pred == 0),
             FP = sum(obs == 0 & pred == 1),
             FN = sum(obs == 1 & pred == 0))
}

accuracy_table(a, b)

This function relies on observations and predictions being coded with 0 as the negative case and 1 as the positive case, so it won't work as desired in other cases.

adviksh
  • 123
  • 6