Editor—Although this paper by Pawlikowska et al is nearly six years old, I read it only two months ago.1 I am surprised that there seem to be no letters or articles referring to it to point out that the analysis is flawed.
The authors report results from a general health questionnaire on a scale of 0 to 36. They provide a histogram for the distribution, which has a mean close to 14. The authors quote the mean scores for men and women as 24.7 and 26.2 respectively. They give confidence intervals for these means and for the difference between them. These means are both above the 90th centile of the distribution of general health questionnaire score that they show. They are clearly impossible.
Their fatigue score is also shown as a histogram. Possible values for observations are between 0 and 33, and the mean is also about 14. The means for men and women are quoted as 24.1 and 25.2, both of which seem to above the 95th centile. Again, they are clearly impossible.
There are several more subtle statistical problems: the histograms with unequal interval sizes shown as the same length on the graph; the statement that with such large numbers the distributions of responses to the fatigue and the general health questionnaires follow a normal distribution (the shape of the distribution is not related to the sample size); the ignoring of the cluster sampling; the use of two different scoring systems for the questionnaires. But the quoting of impossible means should be enough to show that this paper is flawed. Why has nobody noticed, in refereeing, editing, reading the paper (several authors have cited it but seem to accept it uncritically)?
Although two people are acknowledged for help with computing, nobody is acknowledged for help with statistics. The authors used SPSS. Did they include a missing data code without declaring it? A few 999s would produce the means they quote. Does this problem run through all the calculations?
I think the authors should be asked to explain this and, if necessary, carry out a reanalysis, with competent statistical advice. Potentially incorrect conclusions, based on faulty analysis, should not be allowed to remain in the literature to be cited uncritically by others.
References
- 1.Pawlikowska T, Chalder T, Hirsch SR, Wallace P, Wright DJM, Wesley SC. Population based study of fatigue and psychological distress. BMJ. 1994;308:763–766. doi: 10.1136/bmj.308.6931.763. [DOI] [PMC free article] [PubMed] [Google Scholar]