[Tomte]

Prof. Drosten is the chief virologist at Berlin Charité.

I'm confident that somewhere in his medical education the concept of false negatives came up, and I'm also certain that he doesn't word his answers in an interview for the general public to the standards of the nitpicking HN population.

[thu2111]

That's a bit weak though. Right now what the world needs more than anything is precise, clear, accurate information. It's really hard to do that, I got called on making a vaguely worded and misleading statement here on HN just the other day. But then again it's not my job to make such statements. When it is, I take more care than I do in this little box.

Why can't we ask for more?

[graycat]

> Right now what the world needs more than anything is precise, clear, accurate information.

Right: Among the things we need, good data is a biggie.

Part of that is what I've been addressing here: Test quality, especially for test results reported by the news media.

[graycat]

No, that misses a big point:

(1) Now, in the present context, quite broadly, the testing is important.

(2) The news media reports lots of testing, e.g., from China, South Korea, Italy, the US, etc., e.g., lots of the testing from early on.

(3) IIRC in some important respects, the good tests are super tough to do: E.g., Dr. Brix mentioned in one of the White House task force presentations that the test the US is using involves amplification which likely means the PCR (polymerase chain reaction), which before 1983 by Kary Mullis, was essentially impossible and remains astounding, amazing, and non-trivial, and then analyzing the results of the amplification, likely also non-trivial.

(4) So, we have to suspect that a lot of the tests done early on were not so good. By not so good, we have to mean rates of false positives and/or false negatives too high.

(4) IRCC, Dr. Brix did mention at one of those presentations that they, the US efforts, are willing to use only "approved" tests and that some of the tests submitted had false positives of, IRCC, 60%. Here we are on the way to coin flipping range.

E.g., there is a news story about a guy offering to do tests in his college dorm room. I can't say that that is impossible, but it sounds not so good.

Also for some days, a big news media issue was essentially "Where are the tests? Why can't we all get tested? When will we all get tested? What is the big hangup on the tests? Why do the tests take so long? Where has the Administration messed up? ...."

(5) And of course, it is easy to get a rate of false positives of 0% -- just always report no virus found. Presto, bingo, 0% false positives.

(6) The media has been reporting the results of lots of tests, getting headlines, news stories, eyeballs, attention, creating anxiety, getting ad revenue, etc. -- that is, dirty stuff.

So, we all have to wonder, for a lot of those early tests from out in Wherever Land, what the heck was the quality? Were they really doing PCR? In particular, what the heck were the rates of false positives and false negatives? We just CANNOT just take for granted that the tests done had any meaningful quality. Then we have a tough time using the reported data from such questionable tests to evaluate the transmission rate, etc. of the virus and, thus, plan how and estimate when we might defeat the virus. In this context, important, quite broadly important.

(7) So, very briefly, I mentioned, as Dr. Brix did in one of the presentations, that the rate of false positives is important and even at 0% can be from a silly (technical term, trivial) test.

This mention never claimed that the medical profession is not aware of the rates of false positives and false negatives. And I'm fairly sure that the medical statistics people, some of the best statisticians there are, are quite aware of the now classic Neyman-Pearson result on how, for whatever rate of false positives specified, can get the lowest possible rate of false negatives. I have a relatively general proof from the Hahn decomposition from the Radon-Nikodym theorem (Rudin in Real and Complex Analysis gives the novel von Neumann proof).

And for such things there is a lot more, e.g., the A. Wald work on optimal sequential testing* (a stochastic optimal control problem).

Moreover I have some qualifications in the field since I published peer-reviewed original research, in Information Sciences, on anomaly detection, both multi-variate and distribution-free. And there I used the S. Ulam result tightness that even in the medical statistics community is likely not well known (it's in P. Billingsley, Convergence of Probability Measures). So, I have some technical qualifications to talk about test quality.

But here I'm not trying to address, comment on, inform, educate, or critique the medical profession. Instead:

My real point is, as Dr. Brix also seems to have had in mind, for the news media reporting we just must keep in mind test quality in particular rates of false positives and false negatives.

Simple point. Maybe some people here read in some other intentions and objected to those. Again, simple point: In this context, we have to be careful about test quality, not just within the medical profession but, now, especially for what comes from the news media. Simple point. And in the context, an important point.

It appears that some people have missed this simple point and presumed I was making some other point.

Know of anything wrong with that simple, important point?