Note the effect sizes in past experiments and consider the minimum effect size that would hold policy relevance. To decide which MDES value to use, review the literature – especially published papers on the sector and country. If you are calculating sample size, then you will need to input MDES. LSMS) or data from DIME impact evaluation in same region or country. You may consider using a high-quality nationally representative survey (i.e. You will need to use the best available data to estimate values for each parameter. You will likely never have all the data you need for your exact population of interest when you first compute power calculations. Questions like the first are answered by sample size while questions like the second are answered by MDES. Given that I only have budget to sample 1,000 households, what is the minimum effect size that I will be able to distinguish from a null effect?.Given that I want to be able to statistically distinguish program impact of a 10% change in my outcome of interest, what is the minimum sample size needed?.When deciding which to estimate, consider the research design and constraints of a specific evaluation. To compute MDES, you must input power and sample size.Īs power is typically set at 0.80, power calculations are usually used to compute either sample size or MDES.To compute power, you must input sample size and MDES.To compute sample size, you must input power and MDES.

Power calculations can be used to calculate sample size, power, or MDES. For detailed instructions on implementing power calculations, see Power Calculations in Stata and, as a compliment, Power Calculations in Optimal Design.When choosing sample size, consider the marginal value of added observations or, in other words, the trade-offs between accuracy and cost. A sample that is too small will not permit you to detect a statistically significant effect, while a sample too large may be a waste of limited resources.For clustered samples, you need to also know the intra-cluster correlation and the average cluster size. To calculate sample size with power calculations, you need to know the MDES, the standard deviation of the population outcome, and the Type I and Type II significance levels.The danger of underpowered evaluations by J-PAL details how underpowered calculations can affect study outcomes.To compute MDES, given power and sample size.To compute power, given sample size and MDES.To compute sample size, given power and minimum detectable effect size (MDES).Power calculations can be used in three ways :.6.1 Baseline Correlation of Outcome with Covariates.5 Additional Parameters for Clustered Sampling.

4.2 Standard Deviation of Population Outcome.

Power calculation stata