# Power calculations to get sample size needed to run an experiment

1. Jan 24, 2014

### Charlotte87

Hi everyone!

I am writing a paper in development/experimental economics concerning how household bargaining power (i.e. how much say the husband and wife has about how to spend household income) affect how much money is spent on children. I am going to conduct an experiment and is now in the phase of doing some power calculations in order to see how large sample I need. Unfortunately, I am having some problems doing these power calculations. I understand their purpose and everything, but I do not really see how I should use STATA to do the correct calculations.

I am doing a between-subject design with three treatments. In each treatment the money available will be varying. So, I think I need to do power calculations where I compare proportions since I will be comparing the share of money spent on children in each treatment group.

Even though I have three treatments, I will only compare two-and-two treatments. I will compre treatment 1 with treatment 2, treatment 1 with treatment 3 and treatment 2 with treatment 3. So far I have used the following command in STATA:

sampsi 0.1 0.3, power(0.9)

The significance level is then automatically 0.05, and 0.1 and 0.3 is proportions (i.e here the share of money spent on children). Does this make sense??

I have made a table comparing proportions from 0 to 1 (with a 0.1 intervals) and made a table, but I am still not sure whether my method is correct.

2. May 4, 2014