Answers are grounded in this paper and cite their sources.
Statistical analysis of comparisons bet- ween laboratory methodsPapers reporting comparisons of alternative methods of measuring the same quantity are common in medical journals.The statistical analysis in these papers is often inappropriate so that the conclusions are unjustified.I was first aware of this problem when a previous editor invited me to carry out an assessment of the use of statistics in papers published in the Journal of Clinical Path- ology.I reviewed 76 papers in the issue from January to July 1981.One of the most com- mon errors was the use of the correlation coefficient to assess the agreement between alternative methods of measurement.I found this error in eight papers ( 1%)-in fact, in nearly every paper reporting such a study.Recently I was looking through an issue of the Journal and came across the paper by Lewis et al which described the SSF particle counter as a method of measuring platelet counts.'These authors used regression and correlation to compare the SSF with both a visual reference method and a centrifugation method.Neither of these statistical methods is appropriate for assessing agreement bet- ween methods of measurement.2A quick scan through several issues of the Journal showed that inappropriate analysis remains a common problem in method com- parison studies.'4The most common aproach remains the correlation coefficient.
No takes yet. Share an insight, caveat, or question.
Douglas G. Altman (1991) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: