Checklist for robust methodology and statistical analyses?

Recent editorial in the CORR was about EQUATOR guidelines and the editors write:

Since reporting guidelines rely on consensual scientific principles and evidence-based research, consulting them before beginning a project will tend to cause clinician scientists to consider why an element of a checklist is included; doing so inevitably results in discussion and consideration of alternative approaches and results in methodological choices that improve study design, conduct, and analysis.

Only if we had a checklist for robust and sound statistical analysis and methodology. This checklist could include items such as: no arbitrary categorization of continuous variables, statistically non-significant findings not labeled as evidence of absence, inference is based on confidence intervals, causal assumptions are described when applicable, exploratory hypotheses are transparently reported etc. How about that to improve “study design, conduct, and analysis”?

Leave a Reply

Your email address will not be published. Required fields are marked *