Often times when skimming over an Oman WhatsApp Number List our eyes are drawn to graphics and other dataviz very quickly, and we don’t take the time – or the trouble – to read the rest of the details. However, the same numbers, formatted differently, can sometimes be used to demonstrate one thing and its opposite. Often, the data obtained is poorly sourced. It is not there to simply quote, by way of example, the census organization which provided the data, but rather the framework which allowed the emission of these data: what were the assumptions made to simplify the initial data set? What were the conditions of the experiment? Was the sample studied representative? Was the study a retrospective or prospective study?

The answers to these questions can drastically change your conclusion about this data. Try for example to compare the carbon footprint of 2 companies when one has taken into account the emissions due to the construction and maintenance of its premises while the other has stopped at the cradle-to-analysis. seriousness of the products they provide, or to compare different curves of remissions of Covid cases without specifying the average age of the patients in each case. Although it can sometimes be a lack of rigor on the part of the author of the graph in question, certain biases can sometimes be intentionally induced to lead you to draw certain conclusions more easily than others.

The mines hidden in the data

For example, if the purpose of the article is to show you a very large increase in the variable X over time, nothing better than to represent the values ​​taken by X in a histogram and to truncate the y-axis. The increase in question then appears much more marked than it is in reality, if the bars of the histogram are returned to their real proportions. This distortion of our perception of data is very well described in one of our previous articles that I strongly invite you to read on the manipulation of data in different types of graphics , and from which the graph above is taken.


Once you have best eliminated the sources of bias related to data formatting or bad sourcing, you need to know how to read what the data really has to say, if at all conclusive. They may simply not be sufficiently polarized to come to a hard conclusion. In this case, no conclusion is a good conclusion, a priori more truthful than if you pushed the interpretation where there is none. And in fact, it is not uncommon to be faced with data which is correlated, but which is in no way linked by a cause and effect relationship. However, we tend to jump to conclusions, especially if it would allow us to confirm and consolidate a prior opinion that we had in mind (this is called confirmation bias).

Reasoning errors

With a graph helping you score a point during a heated debate, you are probably not going to spit in the soup, even when the graph looks like this … The other serious error which would be made on this occasion, besides the confirmation bias, would be the confusion between correlation and causality. Indeed, the age of Miss America seems clearly correlated with the number of murders committed by scalding … Correlation? This example may seem exaggerated to you, but we are not immune to this very frequent error in reasoning, and there are many examples available to us in the press (especially during the epidemic) and where we run to conclusions.

For example, if one seeks to compare the effects of such or such new drug, and one observes that for drug A, the mortality rate is higher than for drug B, the error would be to conclude immediately on the harmful side of the drug A. Because by putting the study (completely fictitious and invented for the article) in its context, the A being more effective for the advanced stages of the disease, it was administered in majority to patients in advanced phase which, therefore, had statistically less chance of survival than patients subjected to drug B. The confounding factor here being the progression of the disease in the patient, it influences both the choice of treatment as well as than on mortality results.

Leave a Reply

Your email address will not be published. Required fields are marked *