Comments by "N Marbletoe" (@nmarbletoe8210) on "Pandemic learning from Japan" video.

@wyqtor I agree. IGNORE ANYTHING with CAP LOCK ON. . It is also time to stop feeding trolls. If people sound silly as in "Who cares?" then we may be doing a dis-service to the world by responding. .
5
@elisevillemaire8344 Ok well maybe just one note too many? lol . In my clumsy way I'm just trying to say that a scientific paper is a sort of masterpiece, it really is. But unlike Mozart, a scientific paper benefits from critical review.
3
@JohnL-m2l They need to do RANDOM antibody testing, not like the Tokyo paper. . The Tokyo paper is great, but it is not valid to estimate Tokyo prevalence, because it was not a random sample. . Be sure your govt. officials know they need to do RANDOM INVITATION SURVEYS.
2
Unfortunately the Tokyo study was not 47% among the population, but among the employees of one company. . It is an amazing result, and needs to be followed up with a random sample survey.
2
@JohnL-m2l Yes, exactly! And they should do it by random invitation survey. . It should also be available to individuals, whoever wants one.
2
@RyanHellyer I find that the authors of your linked study agree with my assessment of random samples: "this is a stronger design." . "As shown in Table 1, these studies varied substantially in sampling and recruitment designs. The main issue is whether they can offer a representative picture of the population in the region where they are performed." . Statistical inference offers two ways to get an representative sample of a population: 1) a random sample survey, or 2) a census of every member of the population. . Since random sample studies can be cheap and easy, they should be used much more. In the linked study, 11 of 36 used random samples. . The Tokyo study was amazing, and should be followed up immediately by random invitation surveys. It's the best (only?) way to confirm that result.
2
@elisevillemaire8344 So... what you're saying is Big Salieri should lead, follow, or get out of the way. . de acuerdo. . let the brilliant scientists do their work, free from Big Sal, composing studies to illuminate the noble workings of the universe. perhaps funding should be blind!
2
@DSimonJones Those two numbers are not collected in the same way, so it is not quite correct to just divide one into the other. . Deaths may be close to 100% counted. Cases, however, might be missed half the time, 90% of the time, or even more.
1
@DSimonJones I respectfully disagree. Science tries to get things right. . Ignoring missed cases, if they are a large proportion of cases, would lead to large errors. .
1
@DSimonJones To estimate prevalence, there are two ways: 1) a random sample, or 2) a complete count. . This is a study Iceland did, https://www.nejm.org/doi/full/10.1056/NEJMoa2006100 that uses random invitation surveys. . I used to count birds for a living. Would you like to know how many pigeons you have? There are two ways 1) do a random sample of your acres, or 2) count them all. .
1
@DSimonJones Goodness gracious I am in agreement with everything you say there. . In many countries, this pandemic has been handled with the expertise one would expect if the nation's best minds were replaced by a forgetful hamster.
1
lol
1
Are there particular genes that do differ between e asia and w europe? the overall percent of neanderthal genes doesn't differ much among populations from outside africa, I thought.
1
@mygoogle7396 Aha! Very interesting. . The map on page 5 is especially easy to see the pattern .
1
Looks like they are both doing very well by the numbers. . The deaths per million are about the same this last week.
1
Yes, they need to follow up with a random sample study. .
1
That would make an amazing sci fi movie. It also could be true. . It is especially plausible because it is common for a new disease to evolve to be less deadly.
1
No, nobody thinks that here.
1
If they want to know the population prevalence, they can do a random sample study. This doesn't require huge numbers. . Large numbers do not really improve an estimate, if it's not a true random sample.
1
Yeah, it wasn't a random sample study So the results cannot be extrapolated to a wider population. . They may need to edit their conclusions before submitting. . More importantly, the world needs to buy a large pair of fuzzy dice, and use them. .
1
@RyanHellyer Agree, they don't claim randomness at all -- the sample group was 1877 volunteers from a company with 11 branches in Tokyo. . In general it is not possible to get strong statistical inference from studies that do not sample randomly. . A knowledgeable person might be convinced, even convinced beyond a reasonable doubt, but this would NOT be due to statistics alone. . I'm sure these are good scientists. I will read the full paper and your link before saying more... and thank you for your highly relevant response! .
1
@RyanHellyer First let me say this is a well written study, with remarkable results. . I actually find a problem in the Introduction, more than the Conclusion. INTRO: "Distinguishing which of these factors come into play requires data on the total prevalence of SARS-CoV-2,particularly among the asymptomatic **general population**. To estimate the incidence of COVID-19, serology tests were obtained..." . (emphasis by ** is mine) . The problem is, this is NOT a representative sample of the general population in *methodology**, though it MAY be very close to that in **results.* . The authors say exactly this later in the paper, but it needs to be more clear in the Introduction. . Peer review is designed to be helpful to researchers, more than to simply reject poor papers. This is a great paper. . Even Mozart needed someone to tell him, on occassion, "too many notes."
1
@cakepeace6974 That makes sense. It is not logical to only count COVID deaths if the person was otherwise saintly and immaculate.
1
Was the Ministry's study based on a random sample? . If so I'd be interested in the citation! If not, it is not a valid estimate for the population, so no big surprise if it disagrees with other tests are similarly invalid for that purpose.
1
This study can't be used to estimate that. . 1) It's not a random sample based study, 2) they excluded anybody who had felt ill.
1