Article Text

Download PDFPDF
Power and precision in research
  1. Angie Wade
  1. Correspondence to Dr Angie Wade, Centre for Applied Statistics Courses, Population Policy and Practice Programme, UCL Great Ormond Street Institute of Child Health, 30 Guilford Street, London WC1N 1EH, UK; awade{at}

Statistics from

Are adolescents with constipation more likely to suffer psychological maladjustment? 1

What percentage of Chiari I-type headaches show improvement after foramen magnum decompression (FMD)? 2

Does Bacillus Calmette-Guerin (BCG) vaccination reduce early childhood hospitalisation in Denmark? 3

Is diagnosis of coeliac disease associated with differences in adolescent anthropometry? 4

Does visual feedback affect the rate of chest compressions? 5

These are all questions asked in recent issues of this journal. In each case, the authors collated information from a sample of individuals to yield an answer to their question. Differing study types were used ranging from observational audits and surveys through to randomised parallel and crossover trials. The study designs, participants, settings, sample sizes and key statistics are summarised in table 1.

View this table:
Table 1

Description of the five studies

Despite these differences, the same basic principle is followed for each. A sample of the relevant group of individuals is identified and from observing what happens to this sample, inferences are made about the wider population. The inferences may be beneficial to similar individuals and those involved in their care. For example, clinicians trying to determine whether to perform FMD2 or parents considering the pros and cons of BCG vaccination.3

How well a question is answered by the study depends on how large a sample was studied in conjunction with other factors such as the variability of the measurements and/or event rates. Both researchers and patients intuitively understand that findings based on a larger sample are likely to be more accurate and will have more confidence in results based on a randomised trial of 1000 individuals than if only 10 patients had been recruited. What is less intuitive is that if a treatment is not shown to be effective in a small sample, it may still have benefits. Similarly, we are …

View Full Text


  • Competing interests None declared.

  • Provenance and peer review Commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.