jrnold/cdb90

CDB90 battle dataset

cdb90's Issues

Negative values missing?

Hello,

I've noticed in the description that some variables are supposed to have negative values (e.g., wina, cea, surpa) but they do not.
After looking at the R Code, I see this is being done by the function "tomissing_neg." Is this deliberate?

I spot checked some of the wina values and saw that the ones that are missing should indeed be -1 (defender victory). Am I making a mistake in understanding reformulation?

Explic

The file src-data/CDB90/CDB90.tsv was made by concatenating the csv files in data/M000121 into a single file. But I also made edits directly to the file out of laziness. These included the changes in Helmbold 1995 version CDB91, described on pp 2-4 to 2-5.

To address this

• Convert and concatenate all WKS files into a csv file
• Use csvdiff or similar tool to find differences in the data, and create a patch that can be applied programatically to the original data.
• Create a script to generate the data from the original files. Probably convert WKS to CSV manually and start the script from there.

Fix values of columns war2 and war3 in battles.csv

These should be strings, not numeric.

Fix missing values (convet -9 / -9999 to missing)

Still -9 and -9999 values in some columns.

Date Data

Could be screwing something up here, but the majority of values in Year, Month, Day, Hour are all 9s. From my exposure to the set in the HERO work, I don't think that was originally the case?

Rename columns bdb_*

Rename columns bdb_* in battles.csv to cdb13_*

Should README.rst actually be markdown?