Abstract
Dear Editors,
We read with interest the article recently published OnlineFirst in the Journal of Feline Medicine and Surgery entitled ‘Assessment of menace response in neurologically and ophthalmologically healthy cats’. 1 The aim of the authors was to evaluate menace response (MeR) in cats and to define the most reliable and reproducible mode of examination. For this purpose, 50 cats without history or clinical evidence of neurological or ophthalmological disease were assessed by two examiners. MeR was scored on a scale of 1–5 and then, to assess a more practical option for the clinical setting, a three-level scoring system was applied comparing modes of examination relative to their presence and score. 1
We congratulate the authors for their valuable work, but we have some comments about this study. One of the authors’ purposes was to compare scores between examination modes and the χ2 test was used. This test, however, is used when all variables are nominal, 2 and Kruskal’s gamma or Kruskal–Wallis tests should have been used instead;3,4 Table 1 shows the results of these statistical tests. Also intra- and inter-rater agreement (intrinsic and extrinsic video assessment) was calculated using Cohen’s kappa (κ) test. A κ measure is sensitive to the distribution of the marginal totals. If the prevalence is high, chance agreement is also high and κ is reduced accordingly. The κ value is also dependent on the number of categories, as well as ordinal scoring systems.5–7 For the categories assessed by multiple raters Fleiss κ was introduced.5–8 Also for pairwise comparison of systems, an adjustment of κ for calculating the agreement is necessary. 7 Given these limitations of the κ statistics, the authors should interpret the results with caution.
χ2, Kruskal’s gamma and Kruskal–Walls tests for comparison of an ordinal score system in three examination models
