Good post that calls the official Chinese numbers into question, comparing them to Vietnam-era body counts.
Best evidence: a curve fit several days ago did indeed closely predict reported deaths.
Rest-of-world numbers at JHU are likely more reliable.
epsilontheory.com/body-count/
Best evidence: a curve fit several days ago did indeed closely predict reported deaths.
Rest-of-world numbers at JHU are likely more reliable.
epsilontheory.com/body-count/
You could do a formal statistical test, but well-ascertained organic data would look like scatter around a sigmoid, as @EpsilonTheory notes.
Qualitative reports from on the ground as well as the scale of quarantine suggest significant underreporting.
Qualitative reports from on the ground as well as the scale of quarantine suggest significant underreporting.
My general belief is the official Chinese numbers are loose lower bounds. I don’t think they would *overstate* the crisis.
What about rest of world numbers? Well, they have increased >110X in 20 days, from 4 to 449.
gisanddata.maps.arcgis.com/apps/opsdashbo…
What about rest of world numbers? Well, they have increased >110X in 20 days, from 4 to 449.
gisanddata.maps.arcgis.com/apps/opsdashbo…
Robert Wiblin claims that the numbers actually aren’t being faked. Seems like it should be relatively easy to settle by looking at timestamp of original prediction.
Can’t seem to see his FB post though. @robertwiblin, could you post your analysis?
Can’t seem to see his FB post though. @robertwiblin, could you post your analysis?
1) Here is the original Reddit link with the timestamped prediction from five days ago: reddit.com/r/dataisbeauti…
2) And here is the raw link to the WHO data: who.int/emergencies/di…
Check predictions against reported deaths for yourself.
2) And here is the raw link to the WHO data: who.int/emergencies/di…
Check predictions against reported deaths for yourself.
Relatedly, this post from 2016 makes the case for manipulation of financial stats with an eyebrow raising graph.
There may be an innocent explanation here due to incentives for cutoffs, but would love to see an independent replication of this chart.
seekingalpha.com/article/403153…
There may be an innocent explanation here due to incentives for cutoffs, but would love to see an independent replication of this chart.
seekingalpha.com/article/403153…
Don’t know if AUS comparison is right. It supposes full ascertainment (all deaths reported) & a stationary process (death rate constant over time).
But China seems to have deficit of test kits & death rate varying over at least space (Hubei vs rest). Stats still seem too clean.
But China seems to have deficit of test kits & death rate varying over at least space (Hubei vs rest). Stats still seem too clean.
Like, if you had test kits limiting ascertainment, may see surges as test kits arrive in a region.
Or if death rate varied across time due to new hospital beds, may see declines.
Still feels too clean, IMO. But we are reasoning from small data.
Or if death rate varied across time due to new hospital beds, may see declines.
Still feels too clean, IMO. But we are reasoning from small data.
When evaluating something like this, prior probabilities are important. I would tend to trust Singapore.
But the probability of some within the Chinese state fudging numbers is not zero. Indeed, parts of the Chinese government have recently accused other parts of faking data.
But the probability of some within the Chinese state fudging numbers is not zero. Indeed, parts of the Chinese government have recently accused other parts of faking data.
