rdstern
commented
4 years ago If these arguments are accepted, then it is possible that the work on satellite/reanalysis data, PICSA, and mapping could also be proposed for different components of teaching statistics. So mapping could be included when spatial statistics is taught, PICSA is a nice case study - possibly already covered in the main paper, and Danny's work on making satellite and reanalysis data more easily accessible is also relevant.
I propose that each of these areas (extremes and circular data analysis) could easily be included in our reformed teaching of statistics.
To make the case I suggest it follows easily from our new paper, that suggests the use of climatic data. To make the argument stronger it should also follow from the Cobb-style reforms. @volloholic can we make that point. Of course I would include R-Instat here, but I suggest the paper would be better (and more powerful) if we offered alternatives. Thus, if R is taught first, so it can be assumed by the students, then we can include these subjects just as easily - or even more easily - directly. The extRemes package has its own menu-driven interface, and circular is also a reasonable package.
If so, then the journals on teaching statistics are an obvious place to aim to publish two methodology papers - and largely methodology on teaching, as well as methodology on extremes and circular data analysis.
I suggest the circular follows the standard ways we would like to suggest statistics be taught. So there is a lot to discuss in descriptive statistics and the additions for the interns will be on distributions and modelling - particularly, for me, regression modelling!
With extremes the descriptive point is that the data have usually to be summarised before being modelled. Also our interest is often in estimating quantiles, e.g. return periods, rather than the mean. And we have asymptotic results that parallel the central limit theorem. And the GEV distribution is interesting, because it can be skewed in both directions - so a very flexible family of distributions.
So, for David - how does it relate to Cobb? And for James - how might it be included in the teaching in Maseno? We can discuss including in AIMS teaching, but what about in an u/g and/or MSc course in applied statistics in a University?