|
Field
Last Published
|
Before
June 21, 2024 09:57 AM
|
After
January 27, 2025 03:59 AM
|
|
Field
Primary Outcomes (Explanation)
|
Before
Please refer to the pre-analysis plan which includes the exact survey questions for all outcomes as an appendix.
|
After
Please refer to the pre-analysis plan.
|
|
Field
Planned Number of Observations
|
Before
1,500 households, surveyed once at baseline, midline, and endline as well as fortnightly on certain outcomes for at least one year (T = 26) or longer depending on the availability of funding.
|
After
1,872 households, surveyed once at baseline, midline, and endline as well as fortnightly on certain outcomes for at least one year (T = 26) or longer depending on the availability of funding.
|
|
Field
Sample size (or number of clusters) by treatment arms
|
Before
N = 1,500: 900 treated and 600 control households across 72 household clusters (24 high saturation, 24 low saturation, and 24 control group) and 3 evenly sized cohorts of 300 treated and 200 control households each.
|
After
N = 1,872: 936 treated and 624 pure control households, and 312 spillover control households across 72 household clusters (24 high saturation, 24 low saturation, and 24 control group) and 3 evenly sized cohorts of 312 treated and 208 control households each.
|
|
Field
Power calculation: Minimum Detectable Effect Size for Main Outcomes
|
Before
I conducted extensive power calculations in which I simulate and estimate each of the above specifications 1000 times. For specifications (0) and (3), the study is 80% powered to detect treatment effect sizes of the program of at least 0.15 SDs. The spillover specification for research question (1) is powered to detect treatment effect sizes of at least 0.1 SDs. For specification (2), power is somewhat lower, achieving an 80% probability of detecting an effect size of 0.3 SDs or larger of ranking cohorts against each other.
|
After
I conducted extensive power calculations in which I simulate and estimate each of the above specifications 1000 times and assuming 5% attrition. For specifications (0) and (3), the study is 80% powered to detect treatment effect sizes of the program of at least 0.15 SDs. The spillover specification for research question (1) is powered to detect treatment effect sizes of at least 0.1 SDs. For specification (2), power is somewhat lower, achieving an 80% probability of detecting an effect size of 0.3 SDs or larger of ranking cohorts against each other.
|
|
Field
Keyword(s)
|
Before
Agriculture, Firms And Productivity
|
After
Agriculture, Firms And Productivity, Health
|
|
Field
Secondary Outcomes (Explanation)
|
Before
Please refer to the pre-analysis plan which includes the exact survey questions for all outcomes in the appendix.
|
After
Please refer to the pre-analysis plan.
|