Tuesday, March 27 2018
This meeting lasted just over an hour. Professor Robinson and I began to actually work through the coding for the regressions in Stata. I had tried to begin this process the week before with someone from the CTRL lab, but we ran into a few issues regarding the data sheet itself, and thus could not move forward, and there was nothing substantial to report. So between that last meeting and today’s meeting I have been fixing up those mistakes, which included converting all gross data into per capita numbers, changing the original indicators (Freedom House and PRIO) into bivariate. Once these changes were made I was able to actually work through the regressions with Professor Field. We quickly decided that I would also need to linearly interpolate the data for my dependent variables which only appear every 4-5 years at best (most likely from national censuses). The reason for this, even though it is in effect “making up data” is that without interpolating, given the limited availability of many of the other indicators, my cases drop from over 1200 to 68. The reason for this is that Stata will only run the regression if every line of data for a case is filled. While 68 is still enough cases to perform the statistical tests, I decided that it will still be beneficial to run the regressions on interpolated data also (I will certainly keep the original regressions with the 68 cases, if for no other reason than to compare). Because my DV’s (Gini index and Poverty percentages) change so little, and so linearly, I decided that linear interpolation will bring more validity to the project by way of adding hundreds of additional cases than it detracts from that validity by “making up data.” Regardless of this addition, the meeting went very well. I now understand all of what I need to to perform regressions in Stata with panel data, and a few other tests which can be useful, such as bivariate correlation for my DV’s.
After this meeting I feel confident moving forward that I can finish up the ‘meat’ of the project in the next week. I will interpolate the data and perform those same regressions that I did with my original data. Once this is done, all that is left is the interpret the results and write up the analysis – certainly not a small task, but one that I am familiar with and feel comfortable doing independently. Once this is done I will begin revising all of the sections we have been working on all year, and I will synthesize them into a complete paper.