2.53 "Even if the test has not a very good specificity, so is able to classify the patients without the disease". I'm having a hard time trying to understand that bit. If my understanding is correct, a test which has low specificity wouldn't be able to classify patients without the condition. Am I wrong ?
You are not wrong that a test with lower specificity would not be as good as identifying true false (i.e. test and actual are both negative) than one with a higher specificity. But what she is saying here is in the context of *prevelance.* Example: Test A and Test B each have specificity of 80% meaning on average they correctly identify 80% of those without a disease. But these tests will perform differently on two populations of different disease prevalence. In the example she is using, if Test A is given to a population in which only only 3 of 50 patients have the disease, it will perform better than Test B given to a population in which 47 of 50 patients have the disease, despite both tests having the same specificity. Thank you for coming to my TED talk. I am a dog.
Your box plot is not correct, you are supposed to have Disease/No Disease (reality) on top, and the test to the left. You have it opposite... your math is still right, but this isnt how industry writes out the box plot... because if we use NPV/PPV/Sens/Spec equations on this box plot you have, we would get different answers
hello, it is an amazing tutorial, but you have exchanged the formula of sensitivity with Positive predictive value (PPV) and implies specificity also.
Fantastic overview. Thank you so much
You are very welcome!
2.53 "Even if the test has not a very good specificity, so is able to classify the patients without the disease".
I'm having a hard time trying to understand that bit. If my understanding is correct, a test which has low specificity wouldn't be able to classify patients without the condition. Am I wrong ?
You are not wrong that a test with lower specificity would not be as good as identifying true false (i.e. test and actual are both negative) than one with a higher specificity. But what she is saying here is in the context of *prevelance.*
Example: Test A and Test B each have specificity of 80% meaning on average they correctly identify 80% of those without a disease.
But these tests will perform differently on two populations of different disease prevalence. In the example she is using, if Test A is given to a population in which only only 3 of 50 patients have the disease, it will perform better than Test B given to a population in which 47 of 50 patients have the disease, despite both tests having the same specificity.
Thank you for coming to my TED talk. I am a dog.
Great
Your box plot is not correct, you are supposed to have Disease/No Disease (reality) on top, and the test to the left. You have it opposite... your math is still right, but this isnt how industry writes out the box plot... because if we use NPV/PPV/Sens/Spec equations on this box plot you have, we would get different answers
a beautiful girl and a an excellent lecture