Excel-Based Methods for Power Analysis and Sample Size Calculation

Overview

This course gives the basics of sample size estimation. It will be of use to those embarking on a research project and who are perhaps trying to complete ethics and grant application forms. Both precision and power estimation approaches are covered.

The concepts of sample size estimation are introduced, including when and why this might be necessary. Formulae are presented and discussed. Excel sheets are provided and pre-programmed to facilitate use of the formulae. In the first part of the course, the precision-based approach to sample size estimation is introduced, whereby the numbers are selected to produce a confidence interval for the relevant estimate with a specified width. In the second part, we progress to the more traditional power calculations that aim to give the sample required to determine a specified difference with selected power and significance levels. All terminology will be explained as it is introduced in the sessions.

Sample size estimation is discussed for the following measurements: Means, proportions/ percentages, rates, time to event data, measures of agreement and percentiles.

Learning Outcomes

At the end of the course, delegates should understand the main principles of sample size calculations and be able to perform such as calculation for their own research. In particular, delegates will be able to:

• Understand why sample size calculations are important for research.
• Explain the main principles of statistical inference (e.g. standard errors, confidence intervals) and how this applies to sample size calculations.
• Recognise what type of sample size calculation is appropriate for a given research proposal.
• Gather the necessary inputs / estimates that are required to perform the calculation and understand where these can typically be derived or found (e.g. from previous research papers).
• Understand the concept of ‘power’ and how changing this quantity will affect the resulting sample size.
• Have awareness of other more advanced topics such as how sample size calculations might be adapted in the context of cluster randomised controlled trials.