Genetic studies often fail to take into account the effects of their sample size on the sensitivity of their analyses. This can greatly bias the probability of detection of a particular effect, making it only possible to detect very strong effects but nearly impossible to detect weak ones. This is particularly true in studies using binary characters that have a polygenic basis, necessarily creating lower statistical power. This software was designed to aid investigators in determining the optimal sample size for their applications.
The program is very simple... input the desired sample size, the number of iterations for the bootstrap (typically >1000), the lower proportion, and the higher proportion. The program will output to you the number of iterations in which a significant difference (t>1.96) was observed between two samples given your sample size if the two proportions are true values.
For example, say you are doing QTL studies involving one strain that exhibits a trait 90% of the time and another strain that exhibits the trait only 10% of the time. The "discimination difference" is thus 80%, or 0.8. If you want to know the probability of detecting a QTL that contributes to 10% of the phenotypic difference given a sample size of 200, you can test the likelihood of seeing a difference between 82% (which is 90 - 10% of 0.8) and 90%. The result should be about 40%.
Known bugs: First, make sure you always enter the smaller number first. Second, for some strange reason, Windows 98 can't run this program directly, but you can run it either by creating a shortcut to it or by running it directly from DOS.
If you're interested, the file is below. Please cite the following if you use this software:
Noor, M. A. F. and K. R. Smith. 2000. Recombination, statistical power, and genetic studies of sexual isolation in Drosophila. Journal of Heredity, 91: 99-103. See Abstract.
Program (Power.ZIP), 25K
Link back to Noor lab software page.