A friend of mine, from the Social Policy Department, was looking for some statistical advice for her research and came to me. She had already a very clear idea of what she wanted to do and her questions were pretty much about whether she was doing the right thing to defend her thesis. I did some research and some review of the concepts and told her she was definitely on the right track, we both were very happy with our productive meeting and went for lunch together...
At lunch we met with our friend, from the Economics Department. I don't have anything in particular against economists, in fact in my first university there were loads of them and I happen to make a lot of good friends there. However, I also learned something: There are economists who suck at maths and there are economists who are good at maths. Now, I am not a economist but as I understand it, Economics is the study of the optimum allocation of finite resources. Economics is then, not a trivial subject, and it can easily be seen how maths may be very useful. So it is natural to assume that all economists have to deal with some advanced stats at some point during their studies. I guess that based on this assumption, my friend decided to explain the economist about her ideas during lunch expecting to find an agreement... big mistake!
What happened was more or less the following. My friend explained she had done a survey and from the information gathered she had extracted some factors that were in line with the theoretical studies. Using these factors she wanted to explain a couple of variables so she was planning to do a regression analysis with the factors as the covariates. The first thing the economist said was that that was not a regression. At that point I asked why not. He said that that was not possible without really giving an argument. I rephrased the problem as the fact that we wanted to know how the factors explain the variables and even make a geometrical description of the problem and explained how a regression analysis was not only relevant but also possible and we had everything we need it to perform it. He stood on his position: That is not possible. He said that should not be called a regression since the variables we were using were factors and what were supposed to do was a "reduced rank regression". I told him I was not familiar with that technique, that maybe he could be right so I asked him to explain what a "reduced rank regression" is. He was not able to explain it, and every step he gave towards trying to get out of his problem he came up with some new technical terms: eigenvalues, time stationarity, etc.
Most of the terms he used I was familiar with, but I felt he was just giving away technical terms with the hope that at some point we would admit defeat in the sense that we didn't have the knowledge to tackle the problem. Nevertheless, I did know the terms and never conceived defeat, I looked up for the reduced rank regression technique and found out that we could have actually been able to implement the method, but it was pretty much the same as we were doing without knowing the terminology. This makes me conclude that the economist was able to identify the technique he needed in order to solve the problem but when I showed him the method he was not able to recognise it. He doesn't really know the method, he just knows when to use it.
The second discussion developed as follows. A friend from the Aerospace Engineering department came to a friend, who does Geometry, and me to try to find a solution to a problem. He wanted to know how we could find an object in space given only the distances to an instrument. We mathematicians agreed on the fact that we can't find the object unless we had more than one instrument, in fact four in space, and know the position of these instruments in space. However, the engineer wanted to know if there was a solution if we didn't know the position of the instruments. During one of our meeting another engineer listened to our conversation and decided that approach to us and told us that we were simply stupid. He argued that that problem had been solved ages ago and there were even algorithms to find the locations. I explained to him that I didn't see how that was possible and that I wanted to understand the maths behind the maths behind the problem. He insisted that our request was a stupid one and the only thing he was able to do during his "explanation" was to draw to circles on a sheet of paper but he was not capable to deduce a single equation. At this point I realised he had no idea of what he was talking about. My engineer friend asked him the algorithm then but he said someone else had it and we should ask that person. He also manage to give away another technical term: "Multilateration".
Neither my mathematician friend nor I were familiar with the term multilateration, but when I looked it up I figured out that it is precisely finding the position a point knowing the distance to four other points. Which was the problem that my mathematician friend and I were able to solve in the first place. So once again the angry engineer had an idea of the method needed to solve the problem but when seeing the maths of the method he was not able to recognise it.
So, what's the point of all this ranting? Well I came up with a bunch of key points:
1. One should not get angry if someone doesn't know something. One can politely say that a problem has been solved and if there's a technical term by which a problem is known by the experts you can introduce it.
2. Do not try finishing an argument by throwing technical terms. Using technical terms is ok as long as everybody in the discussion knows them. If not, one should introduce them.
3. If one can't do the maths of a certain technique do say so. There's nothing wrong in not knowing everything, what is vary wrong is to say people are stupid for not knowing that someone else knows it.
4. In general, obey Wheaton's Law: Don't be a dick.
No comments:
Post a Comment