I run a lot of logistic regression models at work. At my previous job, where I used Minitab, I always saw stats on Concordant Pairs, Discordant Pairs, and Ties in the model summary output for Logistic Regression modelling. I tried to look for a function that gives you the same stats for a logistic regression model in R, but wasn’t successful. So, I decided to write my own:
An example of the output from this function follows:
> Concordance(GLM.3)
$concordance
[1] 0.6798001
$num_concordant
[1] 2312
$discordance
[1] 0.3201999
$num_discordant
[1] 1089
$tie_rate
[1] 0
$num_tied
[1] 0
You can see from the above stats that the model I evaluated with the concordance function had a concordance rate of 68%, and a discordance of 32%. So, seemingly evidence of a good model!
