Intended for healthcare professionals

Rapid response to:

Endgames Statistical question

Simple associations

BMJ 2008; 337 doi: https://doi.org/10.1136/bmj.a197 (Published 10 September 2008) Cite this as: BMJ 2008;337:a197

Rapid Response:

need to use graphics

I entirely agree with my statistical colleague.

Although the literal "answer" to the question is correlation
coefficient, the most useful practical answer for the real world of data
analysis is scatterplot before correlation coefficient. This is the
message to send. The scattergram conveys much more information. Owing to
pitfalls of outliers, non-linearity, subgroups it is best to look at the
data first. With widespread access to PC packages these days there is no
excuse. Like p-values versus confidence intervals, this debate belongs to
yesterday.

Competing interests:
None declared

Competing interests: No competing interests

12 September 2008
John A Steward
Director
WCISU 13th Floor Brunel House, Cardiff CF24 0AH