Why COVID19 predictions are going wrong?

Use of statics, mathematical models and simulation for different types of analysis dates back to 1950s. Today, technology advancements in the area of data analytics & artificial intelligence has broadened the applications by multi fold.

The beauty of statistics is that it allows you to create any type of narrative, positive or negative. It's easy to tweak the numbers. Just by changing one assumption you can turn around the results and hence the story.

Since the outbreak of COVID19 I have been going through so many predictions on the spreading, impact, recovery from COVID19 using mathematical modelling, statistical analysis combined with artificial intelligence algorithms!

COVID19 is not just challenging doctors & healthcare professionals but also data scientists. If one carefully studies the predictions made since beginning of March and compare them with the real numbers today most of them are far apart.

Why predictions are going wrong? We have a saying in modelling: ‘Garbage in garbage out’!

The input data that is used for prediction has to be precise. Also, it is essential to consider all the parameters that impact and study the variation of each parameter in the complete range.

In simple terms even if you don’t consider one cause in your study or you haven’t captured one cause with sufficient accuracy the predictions can completely go wrong.

In the case of COVID19, most of the studies I have read have excluded ‘asymptomatic carriers’ or have taken an assumed value. Another thing which is very difficult to model is human behaviour.

We don't have data of asymptotic carriers; it won't be available unless extensive tests are done for a huge population. Also, it's difficult to predict primary & secondary contacts because human behaviour is different from person to person. It depends on several aspects. The mathematical models representing human behaviour will have lot of assumptions & approximation. Hence predictions will not be accurate.

These methodologies work very well in a controlled set-up where one can gather accurate & precise information, feed it to the scientific tools and then analyse and predict the response.

Linear & discrete mathematical equations can't predict the non-linear and continuous ‘nature’! Even the non-linear methods and continuous time models are limited by the availability of accurate real-world data.

Only nature can say when is the end to this pandemic!