To track the progress of the pandemic, you want to know, for example, how many people die each day. So, typically, every day, a public health authority will announce something like: there are 489 new COVID deaths today, May 1. On May 2, they announce 189. 138 on May 3. /2
-
-
Show this thread
-
(These are Georgia's numbers.) Most everyone, Georgia included, puts these data down in their spreadsheet. May 1: 489 deaths. May 2: 189. May 3: 138. Graph these numbers, and they're a mess--all over the place--not truly representing disease dynamics. /3pic.twitter.com/uqERkASpTD
Show this thread -
If you're smart, you can smooth to get rid of some of the problems. If you're not (*cough*
@whitehousecea ), you leave the raw data as is. Most public health agencies do something like this for their deaths-per-day graphs. However, these data are NOT in fact deaths per day! /4pic.twitter.com/1QXjImk6fe
Show this thread -
The data represent something rather different: not the people who died on a given date, but the number of deaths that came to the attention of the public health agency on a given date. It's a subtle, but important, distinction; deaths per day is not the same as "new" deaths. /5
Show this thread -
Each day, a public health authority becomes aware of new deaths; some died yesterday, some the day before, some the day before that, etc. So each new datum is partially new deaths, and partially old deaths--corrections to previous tabulations. /6
Show this thread -
The data we need is deaths per day, not deaths-that-came-to-the-authority's-attention per day. And the proper way to track deaths per day is to tally the newly tabulated deaths on the date on which the person died, not lump them together when the death was tallied. /7
Show this thread -
Day-of-death data is gold. Day-of-tabulation data is much less useful. (Here's a graph comparing the two in Sweden. The difference in quality is obvious.) But most authorities present the latter and not the former... /8pic.twitter.com/VeK48ZUp38
Show this thread -
And because it's easier to record new data than it is to go back and alter/correct old data, crappy day-of-tabulation data seems to be predominant over day-of-death data, even when the latter is available... which, in most cases, it seems not to be! /9
Show this thread -
To sum up: death counts presented by public health authorities aren't what they're purported to be, are much worse than data that analysts *actually* want, & which should be available but generally aren't. Then there are those darn data dashboards... maybe another time. :) /end
Show this thread
End of conversation
New conversation -
-
-
Thank you, for this thread. I'm 68. I thought that there was a lot of sloppiness going on, and your rant kind of confirmed it. Don't they teach exactitude, any more? You know, a reverence for accuracy and doing things the absolutely right way?
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.