Manuel Terrádez Gurrea, financial sector data analyst and teaching collaborator in the Master in Business Intelligence and Big Data at the Open University of Catalonia, talks to us in this article about data science and data science.
Two characteristics of the official information issued so far during the coronavirus crisis can be highlighted: from the beginning it has been argued that decisions are made based on the criteria set by scientific experts, in this case healthcare experts; and every day the incidence figures (infected, deceased and recovered) due to the virus are updated.
Data from science, or information from scientists
The question is: do scientists have a single, infallible criterion? Despite the fact that the majority of the population may think so, in any applied science decisions are made in an environment of uncertainty, and therefore mistakes can be made, even frequently. In addition, the problems that society usually faces are complex and therefore there is room for multiple interpretations: two equally competent and prepared experts may have different visions, and consequently, bet on different strategies to address the situations that arise.
On the other hand, the data available is often unreliable.
Starting with the mortality figures, a key issue in this crisis, to compare the virus with other similar ones such as the flu. The data on causes of death are provided by an official body and guarantees such as the INE, with which independence and reliability are guaranteed. Well, when you work with them thoroughly (as for example here), you discover that they are not as reliable as you might expect. And it is not because of the INE, of course, but rather that the problem lies at the beginning of the chain: in the collection of the data. It is an important question, because until one delves into how to obtain data, which is not always possible, it is not easy to know its reliability.
Of course, this circumstance occurs in multiple areas. I often cite an article that casts doubt on the validity of most nutrition-related studies due to, among other shortcomings, the unreliability of the data on which they are based. Hence, many times people end up confused because one day they are told that a certain food is highly recommended, and the next day that it is not so much. We tend to think that there are obscure reasons behind it (manipulation of data or research financed by interested parties), but in most cases what lies behind is simply a basic problem, in the taking (or the storage, or the treatment) of the data. data.
In conclusion, without good data it is impossible to obtain good results , no matter how brilliant the algorithms used to make the models.
Data science applied to virus incidence
On this aspect, as we have already commented, the basic incidence figures are updated daily, by country or by autonomous community. However, relative data are rarely provided instead of absolute data (as they do here, for example), which allow us to put the numbers in context when comparing them with the total population.
Likewise, the consequences of the virus are discussed in very general terms: for example, we all know that it affects the elderly more, but how much more: twice, ten times more …? Because it’s not the same. It is true that being a recent virus, it is difficult to have certainty regarding these issues, but it does not help that data is not usually offered by age groups, although it is something that has improved in recent days.
Of course, nobody talks in terms of probabilities, which is one of the keys when dealing with risk issues, or using common concepts from survival analysis and from the medical or biological fields such as sensitivity, prevalence, relative risk, etc.
At least there has been much talk about the growth “curve” and its “logarithmic” behavior, comparing those of different countries or different autonomous communities, and it is something that is undoubtedly positive. Nor is the comparability of data from different countries easy, because the criteria used by each to count cases can be very different.
Of course, this does not imply that there are not organizations or individual persons who are not transmitting much more precise and elaborate information, but it must be sought, it is not information that comes through the mass media.
In conclusion, it is almost always better to make decisions based on data, even if they are not entirely accurate, than to do so based purely on unconfirmed opinions , but this should not make us lose sight of critical thinking, in this field as in any other. .
