I mean, say this doctor has a 100% success rate but another doctor has 0%. Those two doctors collectively have a 50% success rate but it you have far better odds with the first doctor than the second
@Cenotaph Nope, say the first doctor did 100 successful cases, the other did 2 successful and 2 failed, then the collective would be (100+2)*100/104 = 98.07%
Of course. My point was only that there is definitely a difference between an individual doctor’s success rate and the overall success rate of a procedure across all doctors, responding to the commment I replied to.
98.07 for the surgery in general but not if you have decided to go to the first doctor. Then the 50% chance of the second doctor doesn’t not come into the equation, assuming surgery is done by the first doctor who is independent of second doctor. Hope that makes more sense.
In a statistical regression model, that would be a variable that encodes a specific individual; although encoding hypothetical (the scientific meaning of that word, not the layperson meaning) attributes of that individual is probably functionally equivalent, more useful, and easier to conduct.
How would you scientifically measure a difference between those two definitions?
I mean, say this doctor has a 100% success rate but another doctor has 0%. Those two doctors collectively have a 50% success rate but it you have far better odds with the first doctor than the second
The two doctors would only have a combined 50% success rate if they perform the same number of surgeries
@Cenotaph Nope, say the first doctor did 100 successful cases, the other did 2 successful and 2 failed, then the collective would be (100+2)*100/104 = 98.07%
So the number of cases would matter.
Of course. My point was only that there is definitely a difference between an individual doctor’s success rate and the overall success rate of a procedure across all doctors, responding to the commment I replied to.
98.07 for the surgery in general but not if you have decided to go to the first doctor. Then the 50% chance of the second doctor doesn’t not come into the equation, assuming surgery is done by the first doctor who is independent of second doctor. Hope that makes more sense.
In a statistical regression model, that would be a variable that encodes a specific individual; although encoding hypothetical (the scientific meaning of that word, not the layperson meaning) attributes of that individual is probably functionally equivalent, more useful, and easier to conduct.
Attributes of the surgeons is not easier, because you need to pick the correct attributes.
Really you just need an indicator variable showing 1 if its data from the surgeon under analysis and zero otherwise.
Then test for that indicator variable being statistically larger than 0.