Brian Dew

Should be contained here: https://wonder.cdc.gov/wonder/sci_data/census/cps/type_txt/cps89sep.asp

These are added.

I removed these variables because some results using them didn't seem right and I didn't have time to fix them. The code to generate them is: ``` def family_rel(df): #...

See #112

This was removed from the 1989-93 notebook: ``` def kids_counter(df): '''Count number of own kids under 5 and under 18''' kids = (df.query('AGE < 18 and PARENT > 0') .groupby(['HHID',...

The edited universe changed. It was previously all households in the sample but eventually became only households with a completed interview. Perhaps this is the cause of the issue.

Thinking about switching this variable to one that assigns on person's person weight as the household weight, and identifies which person as 1.

Should be based on CPS design to the extent possible. For example, take into account the area a person lives in and the size of that area.

http://hamelg.blogspot.com/2015/11/python-for-data-analysis-part-23-point.html https://www.methodsconsultants.com/tutorial/variance-estimation-for-complex-surveys/ https://stats.stackexchange.com/questions/315336/stratified-random-sampling-when-strata-overlap https://stackoverflow.com/questions/36997619/sklearn-stratified-sampling-based-on-a-column/36998108 https://newonlinecourses.science.psu.edu/stat506/node/27/ https://stackoverflow.com/questions/2413522/weighted-standard-deviation-in-numpy Some links above--will need to think about this. Start small, including weights, and then, later, try to work on implementations that take into account...