I don't understand how to get the pre-test prob from the question stem.
do you assume that the proportion of positive tests is the same as the pre-test probability?
thank you
Figuring out the pre-test probability is part of this problem and I don't want to give it away, but no,, I would not assume that the proportion of positive tests is the same as the pre-test probability.
Here's a hint that might clarify the difference. In the 2 x 2 tables we've been making the pretest probability is the left column total over the total N and the proportion of positive tests would be the 1st row total over the total N.
thanks. on this point, I hope you read my class feedback comments which i have included in my final exam
Sorry, now I am confused - does D+ refer to those who actually had the disease (and were thus referable) or those who tested positive? Just want to make sure I am clear on the language here since D+ is called "referable" and what is being needed is pre-test probability of disease.
OK, I think I understand the confusion better now. The "Disease" here is "referable diabetic retinopathy." The "test" is the certainty score produced by the machine learning algorithm. Does that help?
Tom
Got it - so when you says, of 8788 evaluable images, 683 were D+ , D+ refers to those with disease = those who had referable diabetic retinopathy. If that is the case, the problem is clear now.
Thanks,
Nikko