Intended for healthcare professionals

Rapid response to:

Endgames Statistical question

Statistical significance and confidence intervals

BMJ 2009; 339 doi: https://doi.org/10.1136/bmj.b3401 (Published 02 September 2009) Cite this as: BMJ 2009;339:b3401

Rapid Response:

On average a poor choice of words

In this quiz, use of the phrase “on average” in the “correct” answer
c is unusual. In describing these data, no reputable journal would allow
an author to presenting the findings using this sentence.

Firstly, what does “on average” actually mean here? If the comparison
being made is that one point estimate of the sample mean is higher for one
group than the other, why not just say that?

Even though the described study is “population based” the study
cohorts are still samples from the population. With the information
provided and confidence intervals overlapping, it is not reasonable to
imply the means in the population are different. If the means in the
population aren’t different, how can “on average” one be described as
being higher than the other?

If response c were phrased "The point estimate of the mean birth
weight was higher in the cohort of mothers with a non-manual job than
those with a manual job" it would be correct. But as it stands I would
argue the laxity of the language means it is incorrect.

Competing interests:
None declared

Competing interests: No competing interests

03 September 2009
Stephen B Lambert
Medical Epidemiologist
Queensland Paediatric Infectious Diseases Laboratory, Royal Children's Hospital, Herston QLD 4029