03-25-2020, 09:05 PM
I was happy to see that they put a disclaimer on the site you linked.
BC
BC
Evidence that social distancing may be working
I was happy to see that they put a disclaimer on the site you linked.
BC
BC
Maybe I'm asking a stupid question that's already been answered, but how do they adjust for the fact that the denominator may be exploding due to many asymptomatic or mildly symptomatic people checking their temperature "just in case"? Anecdotally among my friends and family, there's probably been more thermometer use in the past two weeks than in the past two years. If what they're measuring is the % atypical fevers, what is the denominator? Raw number of atypical fevers this season over total measured temperatures compared to the same % at this time last season? Or is it % of total measured fevers right now? If the former, it does seem like the probable precipitous drop in all other fever-inducing illnesses due to social distancing will really throw off year-over-year comparisons. But if the latter, a ton of people are taking their temperature who wouldn't ordinarily do so absent other symptoms, and who are presumably driving down the % of above-average fevers measured against the total number of temperature measurements. I guess this statistic feels like it has so little capacity to measure what it's purporting to measure that I'm not sure what its utility is. But maybe I'm missing something?
Maybe I'm asking a stupid question that's already been answered, but how do they adjust for the fact that the denominator may be exploding due to many asymptomatic or mildly symptomatic people checking their temperature "just in case"? Anecdotally among my friends and family, there's probably been more thermometer use in the past two weeks than in the past two years. If what they're measuring is the % atypical fevers, what is the denominator? Raw number of atypical fevers this season over total measured temperatures compared to the same % at this time last season? Or is it % of total measured fevers right now? If the former, it does seem like the probable precipitous drop in all other fever-inducing illnesses due to social distancing will really throw off year-over-year comparisons. But if the latter, a ton of people are taking their temperature who wouldn't ordinarily do so absent other symptoms, and who are presumably driving down the % of above-average fevers measured against the total number of temperature measurements. I guess this statistic feels like it has so little capacity to measure what it's purporting to measure that I'm not sure what its utility is. But maybe I'm missing something?
03-26-2020, 06:51 AM
Paging through the site over the past days my impression has been that there isn't a denominator, it's just the total number of fevers reported. You might expect for this to increase some with the increased hypochondria, but as a parent of a toddler (who has a fancy forehead thermometer someone gave us), fevers aren't that hard to spot without one.
Anyway I don't know that that's the answer because as far as I can tell they don't actually say their methodology anywhere, but the maps I linked don't seem to be "per reading" or "per thermometer" or anything else. Just "fevers", "7-day trend in fevers", and "anomalous fevers".
Anyway I don't know that that's the answer because as far as I can tell they don't actually say their methodology anywhere, but the maps I linked don't seem to be "per reading" or "per thermometer" or anything else. Just "fevers", "7-day trend in fevers", and "anomalous fevers".
Paging through the site over the past days my impression has been that there isn't a denominator, it's just the total number of fevers reported. You might expect for this to increase some with the increased hypochondria, but as a parent of a toddler (who has a fancy forehead thermometer someone gave us), fevers aren't that hard to spot without one.
Anyway I don't know that that's the answer because as far as I can tell they don't actually say their methodology anywhere, but the maps I linked don't seem to be "per reading" or "per thermometer" or anything else. Just "fevers", "7-day trend in fevers", and "anomalous fevers".
Anyway I don't know that that's the answer because as far as I can tell they don't actually say their methodology anywhere, but the maps I linked don't seem to be "per reading" or "per thermometer" or anything else. Just "fevers", "7-day trend in fevers", and "anomalous fevers".
The company should have been more clear to the media that what they are measuring is "influenza-like illness" (ILI), and they do show a decrease in ILI in a county that was an early adopter of social distancing/hand hygiene. Would have been easier to realize that the decrease was more likely a decrease in influenza than in Covid-19. Also not clear that the change is statistically significant for trend.
The numbers "problem" should be accounted for in the statistics: they calculate a 95% probability interval and interpret repeated values outside of that interval as "anomalous." If there are few additional Covid cases on top of many expected influenza cases, the overall value will lay within the 95% boundary and will not be deemed "anomalous."
They should be clearer about how the Y-axis is calculated: one shouldn't have to dig through the details in the technical manual linked on the website (and still not find the answer!).
The numbers "problem" should be accounted for in the statistics: they calculate a 95% probability interval and interpret repeated values outside of that interval as "anomalous." If there are few additional Covid cases on top of many expected influenza cases, the overall value will lay within the 95% boundary and will not be deemed "anomalous."
They should be clearer about how the Y-axis is calculated: one shouldn't have to dig through the details in the technical manual linked on the website (and still not find the answer!).
The company should have been more clear to the media that what they are measuring is "influenza-like illness" (ILI), and they do show a decrease in ILI in a county that was an early adopter of social distancing/hand hygiene. Would have been easier to realize that the decrease was more likely a decrease in influenza than in Covid-19. Also not clear that the change is statistically significant for trend.
The numbers "problem" should be accounted for in the statistics: they calculate a 95% probability interval and interpret repeated values outside of that interval as "anomalous." If there are few additional Covid cases on top of many expected influenza cases, the overall value will lay within the 95% boundary and will not be deemed "anomalous."
They should be clearer about how the Y-axis is calculated: one shouldn't have to dig through the details in the technical manual linked on the website (and still not find the answer!).
The numbers "problem" should be accounted for in the statistics: they calculate a 95% probability interval and interpret repeated values outside of that interval as "anomalous." If there are few additional Covid cases on top of many expected influenza cases, the overall value will lay within the 95% boundary and will not be deemed "anomalous."
They should be clearer about how the Y-axis is calculated: one shouldn't have to dig through the details in the technical manual linked on the website (and still not find the answer!).
« Next Oldest | Next Newest »