C1_W2_Lab_1 - errors

Course Q&A AI For Medicine AI for Medical Diagnosis
1

I believe there is an error in the definition of TPR. It should be about TP, not FP, right?
It is how many Positives we predicted correctly out of ALL positives. So the right formula is TPR = TP/P = TP / (TP + FN) and not TPR=FP/FP+TN as you have it.

Moreover the ROC plot in the cell 14 is also labeled incorrectly. The axes have FPR = sensitivity for X axis and TPR = (1 - specificity) for Y axis. So essentially Specificity and Sensitivity are swapped. Which later leads to confusion with the quiz when one of the questions is asking:
“For every specificity as we vary the threshold, the sensitivity of model 1 is at least as high as model 2…”

2

Hi @alexkotlik,

Can you share the screenshot of the TPR formula you are talking about ? I can’t seem to find where it is written as TPR=FP/FP+TN

Also, there are many graphs in the notebook, could you also share the screenshot of the graph you are talking about ?

Thanks,
Mubsi

3

Hi @Mubsi ,

Screenshot 2022-10-12 091150
Screenshot 2022-10-12 0911501710×1188 269 KB

Screenshot 2022-10-12 091230
Screenshot 2022-10-12 0912301701×1171 224 KB

Sure, here it is.

4

Thanks for sharing these @alexkotlik, we shall take a look.

5

@Mubsi I just did this lab, and the incorrect definition is still there. Quite confusing as its not merely a simple typo, but an important metric definition

6

I agree, I was very confused while doing this lab.

7

I can also see the error was identified 4 months ago and I am doing the labs now and it is still not fixed

8

Hi @alice8217 & @Eric5,

I shall check. Thanks for letting me know. This should have been fixed.

Best,
Mubsi

9

This has been fixed.

10

Thanks Mubsi, I just had another look
Appears the error still there
image