Article Text

other Versions

PDF
Which statistical tests should I use?
  1. Allison Shorten1,
  2. Brett Shorten2
  1. 1Yale School of Nursing, Yale University, New Haven, Connecticut, USA
  2. 2Informed Health Choices Trust, New South Wales, Australia
  1. Correspondence to: Dr Allison Shorten
    , Yale School of Nursing, Yale University, Yale West Campus, 400 West Campus Drive, Orange, CT 06477, USA; allison.shorten{at}yale.edu

Statistics from Altmetric.com

Statistical tests can be powerful tools for researchers. They provide valuable evidence from which we make decisions about the significance or robustness of research findings. Statistical tests are a critical part of the answers to our research questions and ultimately determine how confident we can be in the evidence to inform clinical practice. In addition to analysing data to answer research questions, readers of research also need to understand the underlying principles of common statistical tests. It is helpful to know whether statistical tests have been applied in the right way, at the right time with the right data.

There are dozens of statistical tests for researchers to choose from. The statistical tests we use help us gather evidence on which we will either accept or reject our stated null hypotheses1 and therefore make conclusions about the findings of an experiment or exploration. The best way for researchers to ensure they are using the right statistical tests is to consult a specialist in statistics when research is being planned and before data is collected. Mapping out the analysis is an important step in research planning. Statistical tests should not be used as a substitute for good research design or to attempt to correct serious flaws in data.

We will use a hypothetical example to outline some common statistical tests we could use to answer common research questions. Let's say we developed a new ‘mobile phone app’ for patients recently diagnosed with diabetes, with the aim improving patient knowledge about healthy food choices. A sample of patients attending a diabetic clinic was carefully selected according to sound principles2 and 200 patients were randomised into two groups. One group (n=100) would use the ‘mobile phone app’ over a period of 3 months and the other group would be the control (n=100), …

View Full Text

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.