search

ISAAKiel / quantAAR

R Package - Quantitative Analysis in Archaeology
13 stars 1 forks source link

Purpose of this package #8

Open nevrome opened 5 years ago

nevrome commented 5 years ago

I think it's time to decide what we want to do with this package. When we started this 2 years ago we had a lot of ideas, but no concrete plan and therefore quantAAR remained a lazy dumping ground for old code. Some of the ideas we had back then became reality in the meantime (e.g. ISAAKiel/oxcAAR, nfrerebeau/tabula) and others remained unsolved.

I suggest to focus with this package on tidy wrappers and utility functions for the most important multivariate analysis methods in archaeology. I did some work towards this idea and want to hear your opinion. What I want to do:

  1. Closing all current issues (#2, #3, #4, #7)
  2. Cleaning and better documentation + tests + README of the currently available utility functions for correspondence analysis and seriation.
  3. CRAN release with this small set of features in the following weeks.

I'm especially interested in your thoughts @MartinHinz .

MartinHinz commented 5 years ago

Thank you for taking this up again! I actually closed down #2 and #3, but I think #4 and #7 are relevant for this ongoing discussion and its results. From my perspective, it always should go into this direction, but that was before we discovered the benefits of individual packages for individual tasks.

I currently actually am inclined to follow the Unix Philosophy, especially point 1.

I think, what this package can do, is to become a wrapper, as you described. But for doing so, we would need to have a real or made up case study, that a vignette can be build upon. From that we might be able to see, what is actually necessary and desirable.

On the other hand, a lot of functionality I originally had in mind, was already covered by varnastat, that's why I was not so active in developing this idea further (beside the fact that I currently was not so much into cemetery analysis).

nevrome commented 5 years ago

Thank you for this feedback. I agree about #7, because it contains the idea of a nice PCA wrapper which really should be part of this package. #4 on the other hand is highly linked to burial site analysis and therefore leads to yet another package (maybe a modified varnastats). I believe quantAAR should be more generic -- and that fits well to the Rule of Modularity.

A case study is not necessary for v1.0. In my opinion we don't even need example data, because we can rely on the data in the packages we write wrappers for: so far ca and seriation.

My idea here is to keep it very simple and stupid.