Rather than focussing on highlighting the virtues of isolation, certain media outlets have irresponsibly broadcast the results of unverified mathematical simulations, which predict that the number of cases could rise to 10 lakh in India by May. Here’s why they should be trashed.
As the Wuhan virus begins to form local clusters in India, and we enter the second, local transmission stage of this epidemic, citizens and governments have jointly commenced an unprecedented battle to contain its spread.
Thus far, the numbers remain thankfully low, and a surprisingly successful nationwide lockdown is in the process of being put in place.
But fears remain — genuine, valid fears — on how the epidemic might grow.
In such a situation, rather than focussing on highlighting the virtues of isolation, certain media outlets have irresponsibly published and broadcast the results of unverified mathematical simulations, which predict that the number of cases could rise to 10 lakh in India by May.
This has naturally caused panic in some quarters. To counter this, and to deflate needless fears, Swarajya offers multiple reasons why such dire forecasts are unrealistic, and based on an unscientific approach.
By definition, a simulation is a mathematical model constructed to define the characteristics of a phenomenon (in this case, the Wuhan virus epidemic in India), and to then predict its progress. It is a routine exercise employed across disciplines, including fields as diverse as econometrics, reservoir engineering and thermonuclear weapons design, to solve various problems.
At its heart lie a few basic components and rules:
- The static model, which defines in mathematical terms, the environment within which the phenomenon manifests itself (in this case, the geography and demographics of India)
- The dynamic model, which predicts how the phenomenon will evolve (the spread of the epidemic)
- Boundary conditions: the frames of reference within which the simulation is run. In our case, this includes the methodology of epidemic containment (including public isolation efforts), transmissibility of virus, susceptibility, and environmental changes.
- Sensitivity tests: these are predictive runs made to ensure that the input parameters used are indeed representative of the phenomenon being studied.
- Capturing heterogeneity. This is the toughest part, to build a mathematical model which satisfactorily represents the wide variations in human demographics, the spread of the disease, and the physico-geographical environment.
There is more — much more, but their enumeration would lead to death by boredom. So suffice now to say that the main problem lies with the variables. It is their misuse, either willfully, or out of ignorance, which leads to wild, alarmist clairvoyance of the ‘ten lakh cases by May’ variety. Let us look at a few:
Take the Wuhan virus itself for starters. Many believe that this Rakshasa can be killed by Agni; that the virus survives less easily in warm weather. In scientific terms, we would say that an inverse correlation exists between the ambient temperature, and the virus’s ability to reproduce.
The ‘ten lakh cases by May’ brigade confidently reject this conventional wisdom. Based on a study of available information, they say with an extremely high degree of confidence that no such correlation exists. Any Indian reading this would naturally become severely alarmed, and lose hope.
But what does the scientific research actually say? Well, Amitabh Bachchan might have butchered English in Namak Halal, but the authors of the only paper on this subject have been kinder to that tongue. They prove without ambiguity, that such a correlation does indeed exist.
In addition, they also inferred that the virus dies faster with rising humidity. Heartening for Indians, the authors only studied data up to 20 Celsius (because that was the max temperature of the Chinese data sets they used).
As a result, they were unable to demonstrate the increased weakening of the virus, as humidity and temperatures rose further.
That’s encouraging because here, as every Indian knows, two things happen in April: one, it grows so murderously warm that you can fry an egg on the bonnet of a jeep. And second, it grows humid enough in many parts of our land, for frenzied fantasizing of gin-and-tonic on tap.
The adult population of Madras stands ready to adduce themselves as evidence of this! What this means is that whether the ‘ten-lakh-cases-by-May’ tribe like it or not, and in direct contradiction of their assumption, there will be a qualitative, beneficial impact as temperatures and humidity rise across the subcontinent.
A second reason to be highly sceptical of alarmist projections is the non-use of key variables. One factor no one has considered is ultraviolet radiation (UV), which is set to rise as the sun journeys north from the equator to the tropic of Cancer. For decades, fears of global warming were centered on increased exposure to UV, especially during summers. Australia was one of the first countries to demonstrate a correlation between UV and skin cancer.
But surprisingly, these questionable simulations haven’t input UV into their model, even when decades of research show a deleterious effect of UV on living things.
A third reason is the inordinately large spread in projections of cases by May — from a low of 50,000 to a high of ten lakhs. Such a wide outcome spectrum points to a model run with loose controls and wide assumptions.
To invoke a Bollywood analogy, this is akin to ascribing the same oomph factor to Tun Tun and Zeenat Aman (or Keshto Mukherjee and Rajeev Khandelwal, if one wishes to ensure gender parity), while casting for a big ticket, item-number-driven masala blockbuster.
A fourth reason is the non-applicability of analogies to construct models; meaning, that in the absence of adequate data, analysts often resort to similar situations to build their models.
It is an acceptable technique, except that in the case of this current epidemic, there are no analogies applicable to India. The reason is that the manner in which the virus spread in China or Italy for example, is very different from how it spread in India before awareness grew.
Also, the Indian static model is different (demographically, environmentally, etc.). Thus, when doomsayers who forecast a frightfully large number of cases by May, say that they do so on the basis of analogy which is inherently non-representative, that is in itself a negation of the simulation process — and their results.
In a sense, it is a bit like saying that Banganapalle, Kesar, Langda and Alphonso are the same, merely because they all belong to the Mango family. (Brave readers may test this theory by boldly declaring to self-respecting Andhra-ites, that they rate an Alphonso well above a Banganapalle. Swarajya accepts no injury liability on outcomes thereon)
A fifth reason is the absence of history matching. This is a process by which forecast results are overlain on historical data (cases, or deaths, per day per region), to ensure that deviations, if any at all, are within acceptable limits. It provides stringent control on forecasts, and simultaneously offers simple, graphic proof that the workflow employed is logical.
Unfortunately, the forecasts in question have presented no such workings, and thus end up offering more alarm than analysis. A simple plot below explains the importance of this:
As can be seen, even a brief one-week forecast carries a 60 per cent error margin based on quantum of data selected.
What all this means is that Indians are getting unnecessarily alarmed by predictions of a scary spurt in Wuhan virus cases. They must not succumb to their fears, because the sources of these large numbers are obscure, publicly-undisclosed workings, with no representative, clearly-defined models, no control on simulations run, no comprehensive list of input parameters and variables, no demarcated boundary conditions, no recognition of heterogeneity, no history matching on past data, no capture of ongoing containment protocols, and no archetypal analogy to apply.
Indeed, one of the reports actually admits that their prediction of rising cases has been done based on early time Indian data (when the number of cases was miniscule), which showed an estimation mismatch when tested against American and Italian models! That’s techno-babble for a royal foul up.
In conclusion, it may be appreciated that this epidemic will only be stopped by the joint efforts of our state and central governments, and we, the people. There is no place in this fight for baseless forecasts or fear.
Let the alarmists augur not ten lakh cases by May but fifty; it makes no difference. Let them offer more portentous predictions made by vaunted occidental oracles, economists, proctologists or lepidopterists; no problem.
India’s focus will remain on the eye of the fish — social distancing, preventive sanitation, and isolation. That’s all that matters.