Sample Size Power Estimation

Though a necessary accompanier of every submitted research protocol/synopsis, sample size calculations are poorly estimated or reported by many researchers. Indeed, sample size estimation is a logical approach to decide appropriate subject burden in a given study so that there is an optimal trade-off between study resources/logistics on one hand and study objectives/targets on the other hand.

Correct sample size estimation easily empowers a researcher to achieve primary study outcome within a pre-specified type I (alpha or false positive) and type II (beta or false negative) error rates.

Sample size computation is a vast canvas with myriad of formulae depending upon study design, sample selection, allocation, and expected effect size as well as attrition rate on follow up. This can be at times overwhelming for researcher and expert guidance is often sought as shown below:

  • In a cross-sectional study planned to study point prevalence of smoking in college students, such a study design will require at least two pieces of information- a wise guess of expected prevalence and degree of precision desired around prevalence estimate. These two measures may suffice if subject selection is simple random sampling.
  • On the other hand, multistage random sampling may need further information like design effect to accurately calculate desired sample size.
  • In addition, a simple randomized controlled trial (RCT) for obesity control in adolescent students (specialized exercise program versus routine counseling) will need setting up cut-offs for likelihood of false positive as well as false negative error rates along with expected minimum difference in primary outcome (say e.g. BMI or weight between two randomized groups) also called as effect size.