SashaWeinstein
commented
2 years ago Add medians and Re-order columns
I opened up this PR when the work was 50% done, that was my bad.
Add medians
Adding the medians was pretty simple, median wages have leading "MW" in indicator label. Curious that source data doesn't have median wages for occupations crosstabbed by race. It does have median wages by industry crosstabbed by race.
Re-order coumns
The existing order_aggregated_columns worked for the count indicators. It's not exactly right as we would rather have each occupation act as an indicator rather than a category. The way this orders the columns all occupations are together and then all crosstabs on all occupations follow. It would be better to put each occupation crosstab next to the numbers for all races of that occupation but it's ok for now.
I had to write new code for the medians re-order as one indicator is not crosstabbed by race which is an unusual pattern.
I checked the math for the expected number of columns and have assertions that all cols are preserved when we reindex.
Future work
Lots of good refactoring to do for the reordering columns and dcp population processing in general. The assertions to make sure columns aren't dropped is a good QAQC measure.
Unsure
I set the indicator label for income bands as households_<income band>. Easy to change if we want something different
mbh329
td928
Economic indicators for 2008-2012 and 2015-2019
Data comes from DCP population in spreadsheets formatted in similar way to their data for demographics. We have a pretty well-established process for transforming this type of data.
My code to rename cols is slightly different than what's in the demographics code. I wrote a
convert_col_labelscript that gets called for each column label. I also wrote new mappers as the existing ones either include underscores or are specific to certain years. I prefer to add underscores via the logic in the function rather than get them from a mapper