[manwe150]

If I'm reading that chart right, it is the percentage of sicknesses caused by each virus type, not the rate of sickness in the population. So we don't have information in this chart on the absolute prevalence. The change in relative rates may suggest which measures are relatively more effective on different viruses, but doesn't help us analyze the effectiveness of the current narrative.

On the research presented in the original article, what predicted implications would this research have had for March 2020? I thought 2020 seemed like a fairly average year for colds up until March, when COVID spiked. Is that suggesting this benefit/effect weaker in a population than the test measures in a individual, or just that March could have been much worse across the world? (for example, was hypothetically Italy having a very low rate of colds at the time?)

[BoiledCabbage]

> If I'm reading that chart right, it is the percentage of sicknesses caused by each virus type, not the rate of sickness in the population. So we don't have information in this chart on the absolute prevalence. The change in relative rates may suggest which measures are relatively more effective on different viruses, but doesn't help us analyze the effectiveness of the current narrative.

That's exactly correct, and instead the poster is continuously misinterpreting a rate as a fixed quantity.

[timr]

> the poster is continuously misinterpreting a rate as a fixed quantity.

I am not. It is a measurement of the percentage of samples in a surveillance network, positive for each pathogen. Here is the methods paper:

https://publichealth.jmir.org/2018/3/e59/

Here is the relevant section:

> To calculate the pathogen detection rate (as displayed in Figure 2 [second data view] and on the Trend website), we compute the rate for each organism at each institution as a centered 3-week moving average. To adjust for the capacity differences between sites, a national aggregate is calculated as the unweighted average of individual site rates. Only data from sites contributing more than 30 tests per week is included to avoid noise from small numbers of tests. Because the calculation of pathogen detection rate includes results from patients with multiple detections, the detection rate for all organisms can, in theory, add up to greater than one. In practice, this does not occur.

The data says exactly what I described: influenza prevalence has declined to nearly zero. Rhinovirus has continued to circulate at normal levels.

[BoiledCabbage]

I think this will be my final reply on this topic. I appreciate you posting links to support your view, but I don't believe you're taking time to reflect on what I'm saying.

> The fact that rhinovirus continues to circulate so widely

This statement cannot be made from the link you posted. This is a rate. A frequency rate among sampled data. However it does not track the number of samples occurring. No amount of averaging can cover that. As a result, a person cannot make a conclusion to support the above statement on frequency of circulation of rhinovirus.

> Rhinovirus has continued to circulate at normal levels.

This statement is not supported by the data you list. The description you provide above even supports my argument. They are adjusting for capacity differences, but are still reporting a rate.

This is the key statement they make supporting my point:

> To adjust for the capacity differences between sites, a national aggregate is calculated as the unweighted average of individual site rates.

They are averaging rates of infection. You cannot use that to make the statement that you're making that "Rhinovirus has continued to circulate at normal levels."

This is all very well understood stats. The only thing I can suggest at this point is to do some background reading on statistics, sampling and statistical inference.

I'm trying to both be cordial about it, as you have been cordial in your tone and appear to have an earnest interest in this topic - but also trying to make clear that you are unintentionally spreading false information.

You are doing something that statistics 101 makes clear is invalid to do. I'm not certain how else to put it. My only request is that you refrain from repeatedly posting. It is incorrect information, and it turns out it isn't just an invalid conclusion, it's actually the opposite of what is occurring.

Keep the enthusiasm, but just understand the math/stats a bit more.

[timr]

> This statement cannot be made from the link you posted. This is a rate. A frequency rate among sampled data. However it does not track the number of samples occurring.

It does. Click the link [1]. Figure 2 [2] shows the number of samples. They explain it clearly:

> The FilmArray RP test utilization rate (TUR) metric is defined as the non-normalized number of RP patient test results generated each week across the Trend sites (computed as a centered 3-week moving average).

"Non-normalized number of RP patient test results" => count of samples

> To calculate the pathogen detection rate (as displayed in Figure 2 [second data view] and on the Trend website), we compute the rate for each organism at each institution as a centered 3-week moving average

They calculate positive rates for each pathogen, using the the number of samples as the denominator.

> They are averaging rates of infection.

They are not. They are computing an unweighted average rate across sites. Look at figure 2. Read the text again.

This surveillance data is showing you that the rate of samples positive for rhinovirus in their network is ~unchanged. The rate of influenza has disproportionately declined. There is nothing wrong with the data.

> Keep the enthusiasm, but just understand the math/stats a bit more.

Thanks for the advice.

[1] https://publichealth.jmir.org/2018/3/e59/

[2] https://publichealth.jmir.org/2018/3/e59/#figure2

[manwe150]

Your statements seem mostly correct to me, but may be read in a misleading manner to imply the prevalence of colds is similar this year to past years. That data also does exist somewhere (since it was used to compute the plot in your first link), it just isn’t shown in that particular plot. From comparison to the paper, we seem to only have plot 1 for 2020, and not plot 2. Plot 1 (rates) showed the relative effects, but we need a link for plot 2 (counts) for 2020 before we can make statements about the absolute effects.

Hope that helps clear up the misunderstanding. And if you have a link to the counts data also, do share!

[postalrat]

If that's exactly correct why doesn't it always add up to 100%?

[manwe150]

I’ve only spent a couple minutes looking at the plot, but it appears they don’t know why half the patients are showing symptoms in a normal year. This year, something new that they didn’t add to their test aggregation—I assume COVID—increased the “other” category and decreased the “flu” category. But note that means ~80% of the data is categorized as missing last year, so extrapolate this observation at your own risk.