Sample size selection from pilot surveys

Sample size calculations for surveys require knowledge of the variance of the variable of interest, and there are standard formulae for setting the sample size for a survey, based on an assumed level of variance.  In statistical surveys we sometimes perform a pilot survey to estimate the variance of the variable of interest, so that we can calculate an appropriate sample size to use for the main study.  The pilot survey gives us a sample variance, but there is still uncertainty in the true variance, so the standard sample size formulae are likely to be slightly misleading.  In this project we derive an alternative sample size formula that incorporates sampling error in the variance estimate from a pilot survey.  We compare this with the standard formula for setting sample size and explain how these differ.