1/THREAD
PREDICTION: #2019nCoV infection rate will fall off prior exponential trajectory being tracked by @chrismartenson and @biancoresearch creating a sense of relief that the crisis probably isn't as bad as it first looked.
This thread details why that will be a FALSE SIGNAL
PREDICTION: #2019nCoV infection rate will fall off prior exponential trajectory being tracked by @chrismartenson and @biancoresearch creating a sense of relief that the crisis probably isn't as bad as it first looked.
This thread details why that will be a FALSE SIGNAL
@chrismartenson @biancoresearch 2/The exponential progression in which the case count was increasing by 25% to 40% PER DAY may CONTINUE, but REPORTED case count from China will be much lower.
REASON: IF the REAL case count continues to grow on prior trajectory, China medical facilities will turn most new.....
REASON: IF the REAL case count continues to grow on prior trajectory, China medical facilities will turn most new.....
@chrismartenson @biancoresearch 3/...cases away. Facilities are overwhelmed and can't handle new cases. This will create the APPEARANCE of the situation improving and make prior exponential growth projections look WRONG.
In reality, all it means is the situation in China is so bad the data are useless.
In reality, all it means is the situation in China is so bad the data are useless.
@chrismartenson @biancoresearch 4/ The solution is for observers to start focusing exclusively on ex-China data, where reporting is more trustworthy.
This will allow careful monitoring of case and death toll trajectory independent of Chinese government distortion. Much cleaner data, but moe importantly,...
This will allow careful monitoring of case and death toll trajectory independent of Chinese government distortion. Much cleaner data, but moe importantly,...
@chrismartenson @biancoresearch 5/...This approach will also eliminate distortions caused by prior flaws in Chinese reporting. And perhaps even more importantly, this overcomes "But there have been no deaths outside China" objection.
We need to start tracking the EX-China dataset as a new, cleaner source...
We need to start tracking the EX-China dataset as a new, cleaner source...
@chrismartenson @biancoresearch 6/...of data for measuring what this epidemic will mean to the rest of the world, ex-China.
BUT we haven't yet gotten past the "inflection point in the knee curve". Ex-china death toll won't be relevant until it starts to grow to meaningful values on its own.
BUT we haven't yet gotten past the "inflection point in the knee curve". Ex-china death toll won't be relevant until it starts to grow to meaningful values on its own.
@chrismartenson @biancoresearch 7/So, we're about to enter a "data blackout" period. Chinese data will rapidly decay to zero statistical value because it will be so distorted by systems being overwhelmed in China that the data becomes useless. Meanwhile, ex-China data set isn't yet big enough to be meaningful.
@chrismartenson @biancoresearch 8/ So what do we do? Answer is start focusing primarily on ex-China data. For a while that will mean looking at a very small case count and trivially small death toll, and feeling tempted to think "No big deal".
What is ABSOLUTELY CRITICAL is that we measure it in terms...
What is ABSOLUTELY CRITICAL is that we measure it in terms...
@chrismartenson @biancoresearch 9/...of EXPONENTIAL TRAJECTORY, not in terms of how significant the early values are.
Example, suppose you see reports of death tolls that grow from 1 to 2 to 4. Human nature is to think those are very small numbers and nothing to worry about. BUT...
Example, suppose you see reports of death tolls that grow from 1 to 2 to 4. Human nature is to think those are very small numbers and nothing to worry about. BUT...
@chrismartenson @biancoresearch 10/...to a trained eye, 1..2..4 is a SCARY set because it shows a 100% per period exponential trajectory. A trained observer knows than in just a few more periods, the figures will be 8,192... 16,384... 32,768... 65,536...
The point is focus on the exponential trajectory.
The point is focus on the exponential trajectory.
@chrismartenson @biancoresearch 11/ The key is to ANTICIPATE that China data are about to become meaningless, and that ex-China data are EXPECTED to stay small in absolute terms for another few weeks.
During this "data blackout" we need to focus on trajectory, not magnitude of ex-China data values.
END THREAD
During this "data blackout" we need to focus on trajectory, not magnitude of ex-China data values.
END THREAD
