Misrepresenting random sampling?: A systematic review of research papers in the Journal of Advanced Nursing

Aim. This paper discusses the theoretical limitations of the use of random sampling and probability theory in the production of a significance level (or P-value) in nursing research. Potential alternatives, in the form of randomization tests, are proposed. Background. Research papers in nursing, medicine and psychology frequently misrepresent their statistical findings, as the P-values reported assume random sampling. In this systematic review of studies published between January 1995 and June 2002 in the Journal of Advanced Nursing, 89 (68%) studies broke this assumption because they used convenience samples or entire populations. As a result, some of the findings may be questionable. Discussion. The key ideas of random sampling and probability theory for statistical testing (for generating a P-value) are outlined. The result of a systematic review of research papers published in the Journal of Advanced Nursing is then presented, showing how frequently random sampling appears to have been misrepresented. Useful alternative techniques that might overcome these limitations are then discussed. Review limitations. This review is limited in scope because it is applied to one journal, and so the findings cannot be generalized to other nursing journals or to nursing research in general. However, it is possible that other nursing journals are also publishing research articles based on the misrepresentation of random sampling. The review is also limited because in several of the articles the sampling method was not completely clearly stated, and in this circumstance a judgment has been made as to the sampling method employed, based on the indications given by author(s). Conclusion. Quantitative researchers in nursing should be very careful that the statistical techniques they use are appropriate for the design and sampling methods of their studies. If the techniques they employ are not appropriate, they run the risk of misinterpreting findings by using inappropriate, unrepresentative and biased samples.