ORIGINAL: Canoerebel
We're talking about two different things.
By "Fudging the numbers," I mean bad faith manipulations to present a false narrative. That's going to be hard to do for anybody under heavy scrutiny, as are most US jurisdictions (and likely many in other nations too).
Good faith variations in how to interpret things isn't what I was talking about.
ORIGINAL: mind_messing
ORIGINAL: Canoerebel
I doubt anyone thinks the numbers anywhere are perfect (I said as much, above). In some places, they are more suspect than others. But the instance you cited is statistically meaningless. It did nothing except suggest to a casual reader that Georgia's numbers are suspect, which isn't true (at least, not based on that instance).
It'll be hard for any widespread fudging of numbers in places where there is scrutiny. Not only would the perpetrators get caught, they'd be crucified in print. Can you imagine what would happen to a governor or to a state health director that said, "Hey, let's play with numbers to make Covid look worse/better than it actually is."? In every case, somebody would blow the whistle. Who would take that chance? It's going to happen here and there, but most instances of bad faith trickery are going to be isolated/limited in scope.
Hey, wherever there are hotspot, those jurisdictions are reacting. Where there aren't, things are proceeding per easing. That's the only way it can work, right?
I can think of any number of ways to fudge the numbers.
Are the state level statistical publications tied to a specific definition to ensure standardized reporting?
What about assigning location of death? Do you record where fatalities lived, or where they were tested?
Or test numbers. What if you give one person six tests over a week. Have you tested six people, or one? Do you count confirmed false positive/negative results?
The situation in the UK was such that the head of the Office of Statistical Regulations got involved by writing to the health minister. I imagine the US, with the granular state system may even be worse.
It's absolutely not hard to do - there would be justification for doing everything I've listed, either from a pragmatic or a statistical standpoint. That justification standing up to challenge is a different story.
However, it's not likely to be challenged because the devil is in the detail.
Your own Georgia DPH is a good example:
https://dph.georgia.gov/covid-19-daily-status-report
You can see it's combining the serology and viral testing for Coivd. That boosts the total tests number.
But it lets you fudge the numbers. That way, you'd be right to claim that from nearly 800k tests, you've a 8% positive rate.
In actual fact, what you've done is taken two tests:
Test 1 tells you if someone currently has Covid, and this test might be conducted several times.
Test 2 tells you if someone has previously had Covid.
Now, test 1 would be your expected metric for determining the %positive.
But, now here's the kicker. Stick the two together and suddenly you can claim to have done nearly 800k tests (versus 663k of test one alone). What's even better, the %positive drops to 8% (versus the 8.4% of test 1 alone).
See, nice and easy. Just need to compare apples with oranges.
; Julia Child
