Today I write about doctors, but not about those who have this label because they “know”, so dotti.
As Giovanni also says in a famous sketch by Aldo Giovanni and Giacomo, the doctor is interested in the symptoms, not the diagnosis. Many patients go to the doctor and tell them”I have x,y,z, searching on Google I saw that the symptoms are compatible with these pathologies”. These situations are so unnerving that some GPs, agreeably, have put up a poster in their office that says “graduated on Google”.
Clearly patients also do this in good faith. I do it too, or rather I did it, in 2023 and before, but not because I have little trust in doctors in general.
It is done in good faith, first of all because we know that the GP is full of patients, works a lot and therefore we want to make this category use as little time as possible.
So even in the case of the statistician, arriving already there with the diagnosis may actually seem like an act of thoughtfulness towards the professional.
If you say a statistician, or data scientist, I would like x with x something technical, this case is already the name of the diagnosis. And by technical x I mean for example linear model or ANOVA: you are wasting both the statistician’s and yourself’s time. A deadweight loss for society, from a game theory perspective.
That technical x generally comes from some Google search, in many cases Google gives you misleading results as in the case of pathologies. Because Google made you read some content written for example by some computer scientists, with incomplete training, something that has already been addressed in various places.
An example of a technical x: “I would like a model with an r squared (or coefficient of determination) of 0.95”. It means a model that explains 95% of the variability of your target variable. There are a lot of dirty tricks to get that optimistic result. I also saw the equivalent of the model that explains binary rather than quantitative variables, “I would like a logistic model with over 95% accuracy.”
When you say I would like x and the statistician immediately does what you say, the risk of having to deal with a novice in the profession who immediately hits the ground running also increases. In fact, it can lead to unpleasant consequences. This also applies to different areas of work.
Or you increase the risk of copy-paste analysis.
Transform that “I would like x” into this: “I would like to understand how these things influence each other, if this relationship exists”. Another example to transform: “I want a model for which customers abandon the recurring purchase of my product/service” which in technical terms we call churn analysis. In reality, it may happen that you need a sales data dashboard, not a model that predicts the probability of abandonment of each customer. On Youtube, I gave a substantial example for a fuel service.
To summarize, bring the symptoms, the data, to the statistician, and they will think about the diagnosis.
Obviously the appropriate statistical tool for therapy may not be enough. You might need, for example, who takes care of ads, content creation, or see how vendors collect data in a CRM (a kind of database for potential clients and customers), since the statistician is not enough for therapy in most cases.
Except in particular cases, such as large companies, the statistician cannot even make a prognosis (I hate Greek terms), in fact the general practitioner refers you to the specialist (dermatologist, neurologist and so on) if they do not want to give you a medicine that does not prevention.
Finally, common sense suggests turning to the doctor, and even the statistician, to do prevention, and not when it’s too late for a cure, also because everyone knows that treatments cost much, much more than prevention. So if you want prevention I’ll wait for you for a free face-to-face session, but not forever.