Health Cohen, Dupas, and Schaner (2015) conducted a randomized controlled trial investigating the impacts of subsidizing Artemisinin Combination Therapy (ACT, the best known treatment for malaria). You can click here for the paper, though you can answer the problem set questions without referencing the paper. The authors randomized individuals into four groups: one control group, and 3 groups with different ACT subsidies (subsidies at 92%, 88%, and 80%). They estimate effects on various outcomes using regressions of the form Y Bo+BT+BT + By + i (1) where TX is an indicator for being assigned a subsidy at level X. Technically, they include control variables as well as treatment indicators, but throughout the problem set, assume the only right-hand-side variables are the treatment indicators. To start, let's investigate this regression specification. 1. How could you get the mean in the control group from the estimates of this regression? 2. How could you get the mean of each of the three treated groups from the estimates of this regression? 3. B will be a treatment effect it will be the effect of being assigned to treatment condition X rather than treatment condition Y. What are X and Y? 4. How could you get the effect of being assigned the highest subsidy (92%) rather than the lowest subsidy (80%) from the estimates of this regression? Next, let's consider baseline balance. Table 1 (below) presents baseline balance tests. You can think of these as estimates of equation (2), where each row is one regression and Y, is a baseline variable (female, age, etc.). Note that regression tables typically present one regression per column, but the convention for balance tables is one regression per row.