Differences in regional coverage between PIP and PCN

[Aeilert]

The result from the regional coverage calculation in db_create_coverage_table() does not match the coverage calculation Marta has done for PCN.

See attached xlsx for identified discrepancies. Highlighted observations where the difference is larger than 2pp and in red those where the difference is >5pp

Largest differences in (comments from Marta S.).:

• LAC in the 80s and in LAC post 2014 (for the latter, maybe something is happening with Venezuela?)
• OHI post 2007- in some years you get a 100% coverage which makes me a bit suspicious as we don’t have data on all countries (e.g., New Zealand)
• MENA in 2002
• Smaller things that come to my mind: Argentina Rural should be treated as missing country (and same for Suriname)

CoverageTable_Feb22.xlsx

TO DO:

• @Aeilert Will create a "a country-year level dataset of coverage from PIP I.e. a dummy variable for each country and each reference year on whether it is covered."
• @ Marta S / @danielmahler : Could you add your Stata code to this issue for reference? Might be good to check for potential code differences, in addition to the data inputs.

cc: @randrescastaneda, @tonyfujs

[Aeilert]

https://github.com/PIP-Technical-Team/pipdm/pull/117 fixes several of the above errors.

Remaining to tackle are:

• LAC 1986: A 1.09 diff in coverage. I think is because HND 1989 is not being correctly counted as covered. Same issue as COL below. Working on this.
• LAC 1989 – 1991: A 9.85 diff in coverage. This is mostly due to COL 1992 not to being correctly counted as covered. We are working on addressing that .
• LAC 1989 - 2005: URY is causing issues (as expected), but not too much because of the small population size compared to the total.
• WLD/TOT: Mostly minor diffs. I think related to the above.

In addition the following are due actual survey differences between PIP and PCN. They will be added to the What's New.

• MNA 2002: A 23.08 diff in coverage. I believe this is due to IRN 2005 having a “datayear” / “survey_year” of 2005.00 in PCN but 2005.23 in PIP (thus not covered in PIP)
• SSA 1997: A 6.99 diff in coverage. I believe this is due to the fact that ZAF 1996 in missing in PIP. ZAF thus doesn’t have coverage for 1997 (survey years are 1993 and 2000.75)
• SSA 1998: A 1.48 diff in coverage. I believe this is due to SEN 2001 having a “datayear” / “survey_year” of 2001.00 in PCN but 2001.31 in PIP (thus not covered in PIP)