Rapid Responses to:

LETTERS: Stephen Westaby, Nicholas Archer, and Neil Wilson Media attack BMJ 2007; 335: 839 [Full text]

Rapid Responses: Submit a response to this article

Rapid Responses published:

Mean results

Tony Waldron   (29 October 2007)

Mean results 29 October 2007
Tony Waldron, consultant physician St Mary's Hospital, W2 1NY. Send response to journal: Re: Mean results	Could we please have an end to authors suggesting that half of all data will be below the mean (mortality rates for cardiac surgery units in this case)? This will be true only if the data follow a normal distribution, and it is highly unlikely that this will be so with the kind of data referred to in the letter of Westaby et al. It is much more likely - I confess that I have not inspected the data myself - that there will be a susbtantial tail either to the right or the left with some units doing much worse (or much better) than others. This will tend either to increase or reduce the mean, respectively, and thus half the units will not be below the mean. In these circumstances, the appropriate summary statistic is the median when the data will be evenly split, and it probable that it is median, rather than mean survival times that should be quoted. Competing interests: None declared