I was just looking at Covid19 data across countries, and something fascinating about it.
Based on what we know about it, if I asked you there is a country with an old population, and then there is a country with young population, which country will see higher deaths?+
Based on what we know about it, if I asked you there is a country with an old population, and then there is a country with young population, which country will see higher deaths?+
+ Your obvious answer would be "older country" right. But when I look at the data as a whole, this relationship does not seem to be strong. Eg. US with median age of 38.3y has death/reported cases rate of 3.9%, but Iran with age of 32.4, is higher at 6.2%?+
+ Now of course this is just one stray instance, right? Or is it? I plotted data for all countries with over 1000 cases reported as on April 12, to reduce the chance that the country has only recently been infected. I made a regression line with+
+ Median age as the independent variable and the death rate as the dependent variable. The chart of 68 countries is below. A R2 of just 3% means there exists no such relationship, not a strong one by far anyway. +
+ Now what happens if I include all the countries for which I have data with no "infected numbers" filter? Here below for 147 countries:
Again, a Rsq of 5.8%, which again means there is no meaningful relationship between age and deaths, just based off this data.
+
Again, a Rsq of 5.8%, which again means there is no meaningful relationship between age and deaths, just based off this data.
+
+ So I wondered "why the heck is this happening?!". My thought then went to the example of Iran vs US. I postulated that 1 way Iran has a higher death rate than USA is if the denominator in the death rate equation for Iran is artifically smaller. Death Rate=Death/Infected cases.+
+ It just means there are a whole bunch of people in Iran who are asymptomatic precisely BECAUSE they are younger, thus are never coming into the net of "infected cases", while death is final and certain. So numerator is ok, but denominator is smaller.
+
+
+ So my thoughts went to my original study where I had demonstrated statistically that "testing testing testing" is NOT the right approach for India and wondered if now that is called into question. You have to challenge your own thesis, after all. +
+ I then plotted another regression line, with Tests per 1000 population as the independent variable, and death rate for each country. Only 48 countries had both data points available. Lets look at the chart in the next tweet.+
+ Now I expected to see an inverse relationship, ie more the tests, lesser the death rate for 2 reasons
1) more tests means more cases detected so deonominator in death rate goes up
2) detection means cure, so numerator in death rate goes down.
Fair enough? But see what happ!+
1) more tests means more cases detected so deonominator in death rate goes up
2) detection means cure, so numerator in death rate goes down.
Fair enough? But see what happ!+
The chart is below:
I can see a negative relationship, but it is certainly not a strong one! A Rsq of just 3%! So higher testing, HAS NOT, resulted in lowering of the death rate to any meaningful extent. Why is this happening?!
+
I can see a negative relationship, but it is certainly not a strong one! A Rsq of just 3%! So higher testing, HAS NOT, resulted in lowering of the death rate to any meaningful extent. Why is this happening?!
+
+ Therefore "Testing^3" doesnt seem to work really. Italy for example tests 2x of Russia, but has death rate which is many multiples of it. There is perhaps another regression that uses both, median age AND testing data, that may provide some answers. but I do not have+
+ the tools to build a multiple regression model. If anyone can it would be great. All the data I have used is available at @OurWorldInData site.
I come back to my original conclusion again. Social distancing, and lockdown is the only workable solution against a virus+
I come back to my original conclusion again. Social distancing, and lockdown is the only workable solution against a virus+
+ that is perhaps engineered to spread like wildfire. The virii's modus operandi is not escaping detection in blood, but to spread like nothing seen before. It is here that it needs to be tackled. As I mentioned in my blog and tweets, +
https://ecopoliticalindia.blogspot.com/2020/04/coronavirus-does-data-justify-mass.html
https://ecopoliticalindia.blogspot.com/2020/04/coronavirus-does-data-justify-mass.html
+ Testing carries 3 fold costs: 1) Economic 2) Scarcity 3) Risk of Contagion. And when in India the detection rate remains 4-5% of the total tests DESPITE testing the most vulnerable population, it is not advisable to follow the testing, testing, testing by-line.+
+ Unless of course another regression equation can be advanced that will show a 90%+ Rsq relationship between testing and cases, or deaths. END.
Addendum: I just noticed that when I include all 147 countries, the relationship between age and deaths reverses. ie Lower age means higher death rate. :)) This ties in to the point that lower age=more asymptomatic=lower denominator. But I dont believe it affects my conclusion.
Addendum 2: I am aware that this model is not exhaustive, and have already said that I am open to an alternative, multi-regression model overruling my thesis. That is how science goes. But worth noting that the "testing" gang has not yet come out with one despite resources.