On the absence of statistical culture

The Covid-19 pandemic, in addition to four million infected and three hundred thousand deaths, has brought with it an explosion of data, models, analyzes, and information. A real outbreak spread mainly online, through social networks that have given birth to dozens of ad hoc websites (including ours). Many data enthusiasts have declared themselves champions of statistical analysis, proposing models and forecasts that, over time, have proven more or less reliable. Although driven by noble intentions, often discordant forecasts and analyzes created a lot of confusion among non-experts, often generating a distorted perception of what was really going on.

The statistical community has started to protest, softly, on the other hand, the majority of statisticians are working at full speed and do not have time to team up or carry on highly visible full-time activities. Thus, the voice of the community remains somewhat muted.

But people are talking about this issue and the response to this protest from the "public" can be summed up in: "But what do statisticians want? Go more on television? They are already everywhere." This sentence sums up the thoughts of some very educated people, highly competent in their field, and of prestige. Sounds a bit like the phrase attributed to Marie Antoinette "are they asking for bread? Give them croissants! " and the part where they say that "they are already everywhere" denotes well the nature of the problem.

In the scientific-technical committees, Data Analysts are recruited more often than statisticians. The fact that the difference is not perceived depends on the total lack of a statistical culture of this country (as many other countries of course). Those who have not studied statistics in a formal or informal way have no idea of ​​the great importance that a simple but central concept has for anyone who studied it: statistics is meant to estimate unobserved quantities, for example predicting future quantities, but more SPECIFICALLY, it is meant to MEASURE the  UNCERTAINTY associated to these predicted quantities.

A simple example again linked to the Covid19. At the beginning of the epidemic, all StatGroup-19 members ventured into predicting the number of infected people from one day to the next, it seemed fundamental to understand how the epidemic curve was made (before knowing the hash behind the data). A method that gave semi-perfect point estimates at the regional level (it was a few dozen wrong even on regions with few contagions) provided huge 95% confidence bands around the curve of contagions, it was plausible to go from 0 to 1 million infected people passing through the exact value. Now I ask who follows us from the first day, have you ever seen those curves? No. We are all statisticians in the StatGroup-19 and a model that produces such uncertain estimates, even if reluctantly, we leave it in the drawer. Others, without even evaluating the uncertainty of their estimates, propose them to decide on everyone's life. But they are not statisticians.