Data analysis code for Arbiter Kenya (aka Cadaster)'s smart assignment study
# arbiter-kenya-smart-assignment-study
Data analysis code for Arbiter Kenya (aka Cadaster)'s smart assignment study
#### Description of the files:
- cleaning.do: Uses the raw data and creates a cleaned dataset used by the other files. It also creates the file issues.xlsx reporting issues in the data.
- descriptive_exhibits.do: Creates tables and graphs that describe several aspects of the data: Number of cases per mediator under different restricions, number of cases referred monthly, case types by year and courtstation.
- Antoine_va_TC.do: First calculates the value added of eligible mediators and then creates two groups, treatment and control, finally it exports a list with mediator id's, VA and their groups in va_groups.csv.
- impact_eval_details.do: Creates tables and graphs relevant to the impact evaluation: number of cases eligible to be in the impact evaluation referred to court monthly, names of the courtstations to be in the impact evaluation, previous caseload of eligible mediators, and number of mediators in treatment and control groups under different conditions.
- power_calcs.do: Reports the power calculations. It does not export any result, the results appear in the screen.
- caseload_TC.do: Compares the previous caseload of mediators in groups T and C and non-experimental mediators.
- Judges_appoitnment.do: Compares the mediators chosen by Judges during the impact evaluation to those recommended by Smart Assignment
- va_shrunk_unshrunk.do: Some comparisions between the shrunk VA and unshrunk VA estimators
- va_TC_diff_unshrunk.do: Compare T and C groups using the shrunk and unshrunk estimators.
- PAP_descriptive.do: Creates descriptive statistics for the PAP
- PAP_powercalcs.do: Power calculations for the PAP
- PAP_simul.do: Power calculations and other results e.g. Brier scores for the PAP.
#### Folder structure:
In the same folder there should be at least these three folders:
- Data_Raw: Where raw data is saved.
- Data_Clean: Where clean data is …