Hacker News Reader

Comment by sigmar | Hacker News Reader

>If a test has 50, 75, 90% false positive rates, as many do

What tests are you referring to? I can't find a reference but I highly doubt the FDA would approve of many tests with that high of false-positive rates.

View on HN

I think these statistics are claiming '50-90% of observed positives are false', not '50-90% of all tests return false positives'. Suppose 0.1% of the population has a condition and the test for the condition has a 0.5% false positive rate. If administered blindly, most positive tests will be false positives.

I think this is the crux of the article's argument - more people testing more frequently and "blindly" means a whole lot more positive results, most of which will be false.

>'50-90% of observed positives are false'

if ^ is what he means, he should write that. The term "false-positive" has a very clear definition in medical statistics.

Fair enough. But the more rare a condition, the more observed positives will be false.

Such a test would still be very useful, when combined with other independent weak tests. If 5 weak tests in a row confirm you have cancer, then you should worry.

And anyway just report the numbers to the doctor and the patient. "This test has a 5:1 likelihood ratio, so if the odds of you having the disease were 1 to 100, it's now 5 : 100 or 1 to 20."

Everyone needs to get a better grasp on probability.

Which doesn't mean it's not a good test to rule out the tested condition if there are symptoms that might indicate the condition. Sometimes it's clinically valuable to rule out a condition, particularly if the appropriate treatment for another possible condition would be drifter if the tested-for condition might be present. There are people whose training makes them, in principle, the appropriate experts to decide when these tests are needed and worthwhile, they're called physicians.

If you read the fine article, you will see a very clear example of how a highly accurate test can still be misleading for clinical practice if the clinician is deciding whether or not to treat for a low-base-rate condition. Examples like this have been in books about statistical reasoning for medical decision-making over and over and over again since the turn of the century.

There's a difference between false positive rate and positive predictive value:

https://en.wikipedia.org/wiki/False_positive_rate

https://en.wikipedia.org/wiki/Positive_and_negative_predicti...

The latter depends on the prevalence of the test condition in the population, which is one of the major points of the OP.

A test can have a low false positive rate but still have a low positive predictive value if the test condition is sufficiently rare (as it is for most diseases). brianwawok was probably referring to tests with a low positive predictive value.

All very true. Perhaps he mistyped, not sure. I'm just pointing out what was written and factually inaccurate.

Your doubt is misplaced.. At least at the high end. Pap smears have a false negative rate of ~20% and a false positive rate of ~10%. Many doctors are now recommending less frequent tests;

http://www.wsj.com/articles/SB125875596169058039

10% false positive rate is significantly lower than the quoted "50, 75, 90%"

How about you read the article the entire discussion is about? You will learn what is meant right there.

Reply on news.ycombinator.com