create new tag
, view all tags, tagging instructions

Proper Interpretation of Results

Lead Author: Peter Bacchetti, PhD

This page provides general conceptual guidance on interpreting results of statistical analyses that are typically used in clinical and translational research, and it links to an interactive tool that provides example text.

Click here to go straight to the interactive tool

A common problem in clincial research is interpretation based only on P-values, most notably the fallacy of interpreting P>0.05 as proving or supporting no effect. Interpretation should also reflect the direction and size of the estimated effect, along with the uncertainty around it as shown by the confidence interval.

Interpretation usually requires assessing whether the estimated effect would be large enough to be important if it turned out to be exactly right. "Important" can mean important for patients, for public health, or for scientific understanding. Similar assessments are also needed for the confidence bounds.

Some good principles to follow when interpreting a study's results include:

  • The estimate is the possible true value that is most supported by the study's results.
  • Values within the 95% confidence interval are reasonably consistent with the study's results.
  • The study provides strong evidence against values outside the 95% confidence interval.
  • Power is irrelevant for interpreting completed studies.
  • Interpretation should characterize the evidence provided about what may be true in general, rather than focusing on statistical significance as if it were an end in itself.
  • Negative statements (e.g., "we found no evidence for ...") tend to be less clear and more likely to be misinterpreted than positive statements of what the study did produce.
  • Plausibility and the existence of explanatory mechanisms for observed effects should also be considered.
  • It may be important to also consider other sources of evidence, such as previous studies or related results.
  • Even when a study provides strong evidence, independent replication is usually important.
The interactive page at InterpretationText gives sample phrasing of primary results, such as would appear in the Conclusion section of an abstract, based on user-provided assessments of the importance of the estimate and its confidence bounds.
Topic revision: r28 - 11 Feb 2012 - 19:13:24 - PeterBacchetti

Copyright & by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
Ideas, requests, problems regarding CTSPedia? Send feedback