todo #1
Description
If you compile this, the figure may not be the correct one because it takes time to print to file.
Still do to:
-
rewrite results section, starting with a confirmatory analysis based on the mock report and then a second part that looks at details of both pre-registered and other outcomes (similar or the 3ie report). This means we need new knitr tables...
-
I am not so happy with the matching analysis anymore. I think it makes more sense to match at the subcounty level, as this is where potential selection bias would occur. At the same time, we already matched before endline data collection, so does it still make sense? Maybe. Try Coarsened Exact Matching at the subcounty level.
-
we should look at those indices more closely. I was in a hurry when I did this for the mock report. We are loosing too many observations. Raffler et all drop from their indices variables with too many missings.