The pandemic is terrible obviously because of the more than 4,000 deaths and over 275,000 infections, and the socioeconomic hurt it brings upon more than 100 million Filipinos (numbers as of September 17, 2020). However, the pandemic is as terrible as much as its statistics are important to scientists.
For example, knowing the number of new positives per day gives us the rate of infection, the number of recoveries per day gives us the rate in which patients heal, and the difference between the two gives us the number of current active patients of Covid-19. Knowing the daily statistics, the said numbers will then enable scientists to track long term trend, roughly conclude whether the numbers are picking up, reaching to peak of the wave, or declining.
But how reliable are the statistics on new positives? Bill Gates as early as May once talked on how reliable the US Covid-19 statistics is, to which he simply called it as “bogus.” He called it “bogus” because the time one is tested to the time one gets the result was three to four days which he argued as too long. The reasoning is that if one is tested on a Sunday, either a person infected can have easily spread the disease to others by Wednesday by the time one finds the positive result. Or alternatively, a person not infected can already capture the disease from others by Wednesday by the time one finds out the result is negative.
How reliable is the Department of Health statistics on new positives? The daily report typically explicitly states that for every 100 new positives, about 80 has been tested the past one to two weeks and that the rest were tested as way back as March. One to two weeks certainly is longer than Gates’ threshold of bogusness. The new positives whose tests were conducted months back must be beyond bogus.
How reliable are the statistics on new recoveries? While this author does not have a measure for bogusness on recoveries, statisticians have some general notion of what is “normal” data. Looking at the past 100 days, the average number of recoveries per day has been 1,816. Some might think of it too big, while some too small. The reason why people’s opinions on the average number varies is because of the wide fluctuation of number of recoveries per day. For example, the smallest number of recoveries per day was 15 recorded on August 7 while the biggest number was 40,377 recorded on Sunday, August 16.
(Digressing in behalf of wonk statistics: calculating for the average which is 1,816 and the standard deviation which is 6,077 and one can say that there is only 2.5 percent chance that the number of recoveries per day can exceed 1,816 + 1.96 ´ 6,077 = 13,400).
What does wonk statistics say? If data is to be described as normal, then the number of days that recoveries exceed 13,400 can be two-and-a-half per 100 days only. It turned out to be double, that is, it exceeded in five over 100 days. Taking out those five anomalous days from data, the average number of recoveries per day becomes 507 while the average for just the five anomalous days is 26,673. Clearly, those five days must be so extreme to distort and push up the entire average to 1,816. This indicates that data, to begin with, is not normal. So, what if it is not “normal”? The fact that data is not normal deems most of the tools used in intermediate statistics textbooks useless. Hence, there will only be limited if any scientific methodologies to conclude whether the numbers are picking up, reaching to peak of the wave, or declining.
What if for some reason there are days that just randomly have excessive number of recoveries? Consider the following. The last four of the five times that the number of recoveries per day exceed 13,400 occurred in the past four Sundays; 23,000 on September 6, 22,000 on August 30, 17,000 on August 23, and 40,000 on August 16. Urban legend says that Sunday is the best day to recover from all the hard work and partying from the rest of the week. Could it be that Covid-19 patients are suddenly recovering on Sundays after getting sick from the rest of the week? That would be silly. The anomalous days falling on Sundays cannot be random.
The only non-Sunday that the number of recoveries exceed 13,400 was on a Friday, July 30. Does it have to do with it being on the eve of Eid al-Adha, the feast of St. Ignatius, or the beginning of weekend? Probably not. Maybe that day was the only legitimate “random” day to have bigger number of recoveries than usual.
The author is the Dean of the John Gokongwei School of Management, Ateneo de Manila University