Download Wham Pow
Wham Pow is a simplified wrapper for the Bayesian R library BEST. Wham Pow can be used to design experiments given a small set of representative data, The function will output a graph showing the expected single direction width of the 95% Highest Density Interval allowing you to decide the most pragmatic sample size for the question at hand.
A quick example: You'd like to know if Skittles and M&M's have different weights. You only have 10 Skittles in front of you now. You weigh the individual candies you have and now you must decide how many more M&M's and Skittles to acquire.
Before we decided how many M&M's to acquire we might ask the question "how big of a difference in weights would we like to know about?" or even "How big of a difference are we comfortable considering as practically equivalent?" If we can answer this question then arriving at a sample size is straight forward. If we assume that the average weight of Skittles and M&M's may be different but the variance around that average is roughly the same we can use the data from the 10 Skittles to decide how many M&M's we would need to decide if Skittles and M&M's have different weights.
All of this is implemented in Wham Pow. We can input the weights of the 10 Skittles we have into the Wham Pow function. Wham Pow then considers how different the average weight of M&M's would have to be in order for us get a statistically significant result. It then does this again and again across a number of sample sizes and returns a graph like the one below. On the x-axis we can see the sample sizes that Wham Pow analyzed, on the y-axis we can see the minimum difference of means required in order to expect a statistically significant result. This difference of means is expressed as a fraction of the sample data mean. So returning to our Skittles vs M&M's example, if we are only interested in determining if Skittles and M&M's are more than %5 different then a sample size of 20 will provide an adequate experiment.
