Nicolas Cage and people drowning in a swimming pool

[arzina544]

In short, a correlation says how likely it is that there is a connection between 2 variables. A causal connection, on the other hand, has drawn a conclusion from this. There is a cause and an effect. Because it is raining, people take their umbrellas with them. Sometimes these assumptions can be logically explained. We then have a clear picture of what the cause and what the effect is, such as with the rain and the use of the umbrella. Unfortunately, it often happens that this cannot be said with certainty. Then it is good to ask yourself whether correlation can lead to causality or whether you can only establish a correlation.

The most important thing here is to think carefully about whether one thing can precede the other and whether there can be no other explanation for the connection. By deducing that it can be logical that there is a causal connection, you will already come a long way towards the right conclusion.


Although statistics tries to eliminate as much chance as possible, it is possible that a significant effect is a coincidence . So it is very important to keep thinking logically. For example, studies show that the more Nicolas Cage films are released in a year, the more people drown in a swimming pool . Do more people drown in a swimming pool because more Cage films are released? Or could more Cage films have been released because more people drown in a swimming pool? Or is the most likely answer to this question: it is simply a coincidence?

Also read: Away with gut feeling, start with data-driven CX [useful framework]
Illustrative image: girl with a swimming cap and swimming goggles.

It is almost impossible to rule out coincidence. new zealand telegram data But it will help you a long way to, in addition to obtaining statistically sound research, to reason well why a connection might exist.

3. Form one story from different conclusions
Every data scientist or analyst would like the data to provide an objective and complete picture of reality. Unfortunately, this is not feasible in practice. There will always be a certain degree of subjectivity of the researcher in the research. For example, two studies can draw different conclusions about the same hypothesis.



This does not mean that either of the two is immediately 'wrong'. Often this means (with the assumption that both studies are legitimate) that both studies reveal part of the truth . This difference can be determined by examining different types of data, but the approach of the study can also play a role.

Descriptive analytics and correlation analytics look at the data differently than machine learning models. And these models also look at the data in different ways. This can therefore provide different insights, but it is the collective conclusions from these different studies that together form the best result. In order to get the best conclusions from multiple studies, it is important to compare the studies and see how they ultimately create one story together .