Statistics Flaws

TO THE EDITOR: It was fantastic to participate in the training run for the new JNMT Twitter-based journal club (@JNMTClub using #JNMTClub). Although only a small number of invited participants were engaged, it promises to be a fantastic medium for nuclear medicine technologists to connect, engage, learn, and participate in continuing education among a global community of colleagues. The format will be of benefit to those experienced in social media and the novice. The JNMT article discussed during the Twitter-based journal club was Dako et al. (1), examining the patients’ perception of quality at the point of care in PET. The Twitter discussion asked the question “Do you think the results of the survey are meaningful?”

The results lack any statistical validity in my opinion. The Likert scale used rendered the data ordinal in nature. The statistical analysis treated the data as though it were continuous data, and this represents a fatal flaw. Ordinal data should not be represented as a mean; the median is the appropriate expression for central tendency (2). Indeed, had the median been cited, it may have given a better insight into any trends. This flaw is easily rectified with appropriate statistical treatment; however, the mean data and associated P values are inappropriate and leave no conclusions that can be reliably drawn. This statistical flaw is not uncommon in the medical literature but represents reputation damage to the research itself, the review process of the journal, the journal, and the broader profession. A working understanding of appropriate use and meaning of common statistical tools is essential for all.