Inconsistant notation

noetsi

Fortran must die
#1
A pet peave of mine in stats. This is a well known author discussing it in the context of time series - but I think you could extend this to cover the entire field. I remain baffled that there is no attempt to standardize even basic usage (how many different notations are there for sum of squares - a dozen?).

It is worth emphasizing that there is still no agreement on notation for exponential smoothing. An appalling variety of notation exists in the literature, and some authors add to the confusion by changing notation from one paper to the next.
http://www.bauer.uh.edu/gardner/Exponential-Smoothing.pdf
 
#2
I think how standardized the notation is depends on how popular or how much the method has been used/written about

especially after it gets in several textbooks, people eventually settle on a notation that tends to get used the most

For example, you almost always see Y = XB + e because this is very old

It may take some time (few academic generations) for something like exponential smoothing notation to settle
 

noetsi

Fortran must die
#3
I don't know, ANOVA has been one of the most popular methods for nearly a century - and it has some of the worse cases of conflicting terms and nomeclature among various authors I have seen.
 

maartenbuis

TS Contributor
#4
A big "problem" is that statistics is not a isolated discipline, but to a large extend a "helper-science" for other disciplines. My impression is that the largest problem is between (sub-(sub-))disciplines, where I have seen such differences lead to real miscommunication. There are also differences within disciplines but they seem to be mostly manageble.

I see real problems in getting the different disciplines to agree on such an excercise. I see a large probablity of such an excercise failing misserably and a high probability that participating in such a process will be a lot of work and involve a lot of very negative reactions. So I will stay away from such an excercise, at least until I have tenure and nothing better to do with my time...
 

noetsi

Fortran must die
#5
That is an interesting point. It also shows, in a different vein, why ANOVA and regression (which are theoretically the same method) are so different. One was developed in agriculture primarily (ANOVA) and regression was not (although I don't know where it did develop, perhaps psychology).
 

Miner

TS Contributor
#6
You also have the problem that there is overlap with mathematics and engineering both of which use their own symbols for identical equations. Even in statistics, many fail to recognize that Greek letters are used for the population and English letters for samples.
 

noetsi

Fortran must die
#7
Yeah. Why they use Greek at all is anoying - I suspect it has to do with the love affair elite british universities had with ancient Greek at the time people like Fischer were developing statistics. Of couse it is also wonderful that excel and word won't automatically copy the symbols, so when you try to past them from articles or books you end up having to manually insert the correct symbols. I commonly end up using the English letters ignoring the issue of population and sample (which rarely applies to my analysis which always deal with populations).

The fact that statistics does not have a governing body, as for example ISO performs with various methods, means there is little hope this problem will be addressed.