better descriptions of datasets

[xrotwang]

Datasets should have a description including the text blob which is used on the sources-page of the web app, but also a bit of info about how the data is coded. E.g. for the Binford data we just came across the fact that there are rows in data.csv which only differ in comment and references:

B1,,1970,B011,30,"Note, first of three estimates provided by Binford (30, 26, 30) for this society","harrison1949:135, 139; urquhart1951:515, 505",Binford_2001_Tables_5.01_and_8.01,"Note, three estimates provided by Binford (30, 26, 30)"
...
B1,,1970,B011,30,"Note, third of three estimates provided by Binford (30, 26, 30) for this society",sellato1994:143-44,Binford_2001_Tables_5.01_and_8.01,"Note, three estimates provided by Binford (30, 26, 30)"

This is surely something that can trip up analyses, thus should be mentioned somewhere.

[kirbykat]

@xrotwang, we should be able to extract all of this info from the paper we published. The case you describe for Binford is something we pitch as a "selling point" for DPLACE (we present de-aggregated data, so users can get a feel for the variance. In this instance, two separate observations provide a value of 30, which might give the user confidence that this is a representative value for the society). But yes, users who have not read the paper would no doubt be mystified, so perhaps I can work on extracting the relevant details from Kirby et al. 2016.

[xrotwang]

Yeah, after writing this comment, it occurred to me that the paper would be the place to describe this. But I still think it would be good to have this info here as well. In particular things that may have bearing on the database schema (like unique constraints) should be mentioned.

[SimonGreenhill]

This could probably go in the metadata.json as text blob? or a DESCRIPTION.md file?