1/10 In statistics, and in general, going back to ‘square one’ is never a bad idea

So I reset w/ respect to COVID and deaths

Again, I follow the ‘heroes of Covid’ @gummibear737 @boriquagato @aginnt @foogatwo @AskeladdenTX @AJKayWriter @NahasNewman @jhaskinscabrera @justin_hart
2/10 Another hero is @EthicalSkeptic who has done great work on excess deaths, so I won’t dive in here.
3/10 Monthly chart of all deaths for all states + Puerto Rico and DC. No normalization or breakouts. Just bifurcation by state. All data goes thru Aug 2020…

My point here is some states (CA is easy to see in pink on top) are trending up long term (as population grows etc)
4/10 Another way to show the same concept of rising deaths (again due to population size, age of population etc, not getting into that here), is just a simple sum of all states
Choppy data in 2019/2020 - probable data issues. Ie, the peaks and lows may not be 100% accurate
5/10 If given a narrow snapshot, you may come to the wrong conclusion…never realizing the uptick death trend is over many years

For example, lumping all together I could claim this is CLEARLY showing way more deaths which MUST be COVID (or anything else I choose to highlight)
6/10 Back to states. NY is interesting. For whatever reason, they are splitting out NY and NYC…

Let me point out, that spike in red covers up an impressive spike in blue…
7/10 Altogether, combining NYC back into NY you get this…which as @Hold2LLC would say….it was BAD
8/10 Other states worth mentioning are IL, MD, NJ and MA. Rather large spikes on all of them
9/10 Other states in the news, TX, AZ, CO, CA and FL. The data is very very choppy, which again, given the larger trend, unexpected (points to data errors that will be smoothed out at some point)
10/10 I leave this thread with some basic thoughts

Don’t be fooled by graph scales
Don’t be fooled by cause / effect bias
Don’t be fooled by simple ratios or other more complex ‘normalizations’
Try to separate the trees from the forest

Devil is in the details
You can follow @DekruifMichael.
Tip: mention @twtextapp on a Twitter thread with the keyword “unroll” to get a link to it.

Latest Threads Unrolled: