Replies to This Discussion

Permalink Reply by RockyRambo on March 7, 2012 at 1:09am

Hi Sharath,

Since sensitivity and (1- specificity) are determined using different cut off points from the confusion matrix, you can get both of them by varying the cutoff points..Let's say I have scores of 1 million observations in my model. If I keep my cutoff value at <  lowest probability (low cutoff) of the model then I won't be able to capture even a single actual negative, because my model is now rating all the data points above the cut off to be predicted 1's and below the cut off to be predicted 0's. Thus my sensitivity becomes 100% and specificity becomes 0% or (1- specificity) becomes 100%. Similarly I can put this cutoff to be more than the highest probability predicted by the model which will give 0% sensitivity and 0% (1-specificity). Thus, I can keep my cutoff value is continuous and I can get different data points for sensitivity and (1-specificity) depending upon what value I keep...Hope this helps,

Varun

• ▶ Reply