Probabilities (General)
by romansh 
, Sunday, April 13, 2014, 01:00 (3876 days ago)
Here's a question what is your immediate response?-Imagine you are in your doctor's office and she says, "well the results have come in from the tests and you may have that ailment we were discussing." You ask, "what is the false positive rate?" and she replies "five percent with no false negatives." You then ask "does that mean there is a ninety five percent chance that I have the ailment?"-What is her correct reply?-If you 'know' the correct answer please hold off anwering for a bit.
Probabilities
by romansh , Thursday, April 17, 2014, 06:05 (3872 days ago) @ romansh
OK so no one hazarded a guess.-Either no one wanted to admit not knowing or they knew. Perhaps they did not care enough to reply. Fair enough.-Well the doctor's reply should be along the lines ...-Well Mr rom you can't come to that conclusion as I have not given you enough information.-So what is the piece of missing information that we need?
Probabilities
by David Turell , Thursday, April 17, 2014, 16:51 (3871 days ago) @ romansh
> Romansh: Either no one wanted to admit not knowing or they knew. Perhaps they did not care enough to reply. Fair enough. > > Well the doctor's reply should be along the lines ... > > Well Mr rom you can't come to that conclusion as I have not given you enough information. > > So what is the piece of missing information that we need?-I admit to knowing litle about statistics, even though I cooked-booked some results for papers that were published, simple T and P values, occasional chi- square..
Probabilities
by romansh , Friday, April 18, 2014, 01:53 (3871 days ago) @ David Turell
I admit to knowing litle about statistics, even though I cooked-booked some results for papers that were published, simple T and P values, occasional chi- square..-That is fair enough David.-Apparently most doctors surveyed are not that familiar either.-The piece of information missing is the incidence of the ailment in the relevant population. -If the good doctor would go on and explain that the incidence of the ailment in "my" population was say one in a thousand then we can calculate what my chances of suffering from said ailment.-Say we took a thousand people (representative of the population) we would expect, on average, one real positive and fifty false postives. -So my actual chances of suffering from said ailment are about one in fifty or two percent and not the ninety five percent I had originally calculated. Obviously the numbers will vary with the type of ailment and test.
Probabilities
by David Turell , Friday, April 18, 2014, 02:50 (3871 days ago) @ romansh
> Romansh: The piece of information missing is the incidence of the ailment in the relevant population. > > If the good doctor would go on and explain that the incidence of the ailment in "my" population was say one in a thousand then we can calculate what my chances of suffering from said ailment. > > Say we took a thousand people (representative of the population) we would expect, on average, one real positive and fifty false postives. > > So my actual chances of suffering from said ailment are about one in fifty or two percent and not the ninety five percent I had originally calculated. Obviously the numbers will vary with the type of ailment and test.-But your comment does not fit my experience. As a practicing physician I might see one of 'these' and one of 'those' rarities in a lifetime. I would have no statistics within my city's patient population available to me to have any way to get to your requirement. I understand your added step adds to the confidence limits of the statistical judgment, but I'm not even sure the Houston health department could help me. On the other hand, in epidemiological studies that type of informaton is invaluable. In my first book I have a section on the political misuse of statistics.
Probabilities
by romansh , Friday, April 18, 2014, 07:11 (3871 days ago) @ David Turell
David -I can't comment how often you would send off for a test whether it would be a biopsy or blood work or some culture analysis etc.--My point is if someone comes back with a positive result and there is a positive false positive to that test. The test itself looses much if not all its meaning if you don't know how to interpret the data.
Probabilities
by David Turell , Friday, April 18, 2014, 16:05 (3870 days ago) @ romansh
Romansh; My point is if someone comes back with a positive result and there is a positive false positive to that test. The test itself looses much if not all its meaning if you don't know how to interpret the data.-I understand that. But 99% of the lab data had known ranges of normal, with other ancillary tests available to follow up on some abnormal result. As an internist, one never relies one one thing. My point is you are bringing up a theoretical problem that can be a real problem in rare cases. I had a patient with a lump on a testicle. Local pathologists were not sure about the biopsy interpretation. It went to the NIH and came back definite carcinoma. Off came the testis. That I remember this case from at least 40 years ago attests to how rare such a situation was. I can think of othr rarities like this from my practice, but there are many ways to reach a definitive diagnosis. A careful history and physical exam will usually get you to the 90% level.
Probabilities
by romansh , Friday, April 18, 2014, 17:27 (3870 days ago) @ David Turell
My point remains David-People in general (and most doctor's apparently) didn't know now to interpret the probability data presented.-I suspect that just about every test a doctor might ask for has a false positive rate associated with it. It might be vanishing small or significant. Anyway it is not point to denigrate the medical profession in any way. I put my life in their hands quite often.-My central point remains people are not always good at handling probabilities.
Probabilities
by David Turell , Friday, April 18, 2014, 20:04 (3870 days ago) @ romansh
> > Romansh:People in general (and most doctor's apparently) didn't know now to interpret the probability data presented. > > I suspect that just about every test a doctor might ask for has a false positive rate associated with it. It might be vanishing small or significant. > Anyway it is not point to denigrate the medical profession in any way. I put my life in their hands quite often. > > My central point remains people are not always good at handling probabilities.-No, we were taught there were false positives and false negatives, but as I've tried to point out, there are alternate ways to come to a diagnosis that cover each other and this problem. You are quite correct that probability is not well understood. Many improbable events happen every day in my life, and I know that. I see them. But if specificity is involved in the event, then the improbabilities disappear.
Probabilities
by romansh , Saturday, April 19, 2014, 00:49 (3870 days ago) @ David Turell
No, we were taught there were false positives and false negatives, but as I've tried to point out, there are alternate ways to come to a diagnosis that cover each other and this problem. You are quite correct that probability is not well understood. Many improbable events happen every day in my life, and I know that. I see them. But if specificity is involved in the event, then the improbabilities disappear.-I never said or meant to suggest doctors are not taught about about false positive and negative results. I fully expect them to be taught about such phenomena. I would also expect them to be taught these things and the how to interpret such data. -It would appear, if the interpretation of false positive and negatives the learnings have not stuck. Of course (I hope) doctors don't rely on a single test for a diagnosis.
Probabilities
by David Turell , Saturday, April 19, 2014, 04:03 (3870 days ago) @ romansh
> Romansh: I never said or meant to suggest doctors are not taught about about false positive and negative results. I fully expect them to be taught about such phenomena. I would also expect them to be taught these things and the how to interpret such data. > > It would appear, if the interpretation of false positive and negatives the learnings have not stuck. Of course (I hope) doctors don't rely on a single test for a diagnosis.-Sometimes that is all it takes: As I have said the average complex diagnosis will have a constellation of factors reviewed and studied. But some diagnoses require just one test. The problem is having to think of the possibility. I had a patient who had enormous bouts of swelling anywhere in the body. Fatal in the larnyx if prolonged. She had been treated with epinephrine by IV and in her thoat. She came to me and luckily I had recently read a case report about "C-1 esterase inhibitor deficiency". At Baylor Med school luckily there was an immunology prof who could run the single test to prove the point which he did. Luckily the research folks studying the condition had devised a treatment to prevent such episodes. -I don't think your worry about probabilities in medical practice is much of an issue. You are looking in from the outside and I don't think you fully understand how or what good doctors are capable of doing diagnostically.
Probabilities
by romansh , Saturday, April 19, 2014, 15:17 (3870 days ago) @ David Turell
Sometimes that is all it takes: As I have said the average complex diagnosis will have a constellation of factors reviewed and studied. But some diagnoses require just one test. The problem is having to think of the possibility. I had a patient who had enormous bouts of swelling anywhere in the body. Fatal in the larnyx if prolonged. She had been treated with epinephrine by IV and in her thoat. She came to me and luckily I had recently read a case report about "C-1 esterase inhibitor deficiency". At Baylor Med school luckily there was an immunology prof who could run the single test to prove the point which he did. Luckily the research folks studying the condition had devised a treatment to prevent such episodes. > > I don't think your worry about probabilities in medical practice is much of an issue. You are looking in from the outside and I don't think you fully understand how or what good doctors are capable of doing diagnostically.-I am not specifically worried about the medical profession not knowing probabilities.-But now you come to mention it ... what is the false positive rate for your Baylor Med School prof's test, and what is the incidence of C-1 esterase inhibitor deficiency in the group that displays the symptoms?-And this phrase here ... who could run the single test to prove the point which he did. worries me. Science is not in the proof business. There is always a degree of uncertainty in science. I hope the medical profession displays some uncertainty too.
Probabilities
by David Turell , Saturday, April 19, 2014, 16:26 (3869 days ago) @ romansh
> Romansh: I am not specifically worried about the medical profession not knowing probabilities. > > But now you come to mention it ... what is the false positive rate for your Baylor Med School prof's test, and what is the incidence of C-1 esterase inhibitor deficiency in the group that displays the symptoms? > > And this phrase here ... who could run the single test to prove the point which he did. worries me. Science is not in the proof business. There is always a degree of uncertainty in science. I hope the medical profession displays some uncertainty too.-You philosophize too much. The professor's test was a measurement of a protein enzyme in blood. There is no false positive. The mechanism of the test and its abililty to achieve an accurate result had been proven. Medicine is in the proof business when trying to establish a diagnosis for a dramatically ill patient. And finally the difference in my outlook and yours is that I know medicine is an art with applied science, not a science in itself. Which I think yhou actually recognize. Mainly deductive reasoning.
Probabilities
by romansh , Saturday, April 19, 2014, 17:45 (3869 days ago) @ David Turell
You philosophize too much. The professor's test was a measurement of a protein enzyme in blood. There is no false positive. The mechanism of the test and its abililty to achieve an accurate result had been proven. -There is no false positive? How do we go about proving this. We can maybe say the rate of false positives is something like <0.000001, implying we have tasted a million people and not found a single false positive.-> Medicine is in the proof business and >I know medicine is an art with applied science, not a science in itself. Which I think yhou actually recognize. Mainly deductive reasoning. ?????-Yes I know medicine has a degree of art to it. That is why I don't think medicine is in the proof business. It is in the business of making the best possible decision within the constraints of time, education, experience, resources, compassion and no doubt many other factors.-Proof is for logicians who have a well defined set of axioms and well defined method of using Boolean operators (and similar).
Probabilities
by David Turell , Saturday, April 19, 2014, 23:16 (3869 days ago) @ romansh
> Romansh;There is no false positive? How do we go about proving this. We can maybe say the rate of false positives is something like <0.000001, implying we have tasted a million people and not found a single false positive.-The test is just a proven way to do a measurement, with a very small range of plus or minus error in the measurement, there is no false negative or positive at issue. Since human proteins are in a normal range, for example blood HDL cholesterol is 30-60. Normal is defined as in the range. > > Romansh: Yes I know medicine has a degree of art to it. That is why I don't think medicine is in the proof business. It is in the business of making the best possible decision within the constraints of time, education, experience, resources, compassion and no doubt many other factors.-Yes, your statement is more accurate, but proof of diagnosis is part of it, to the best of our abilities.