Author + information
- Joshua Fogel, PhD∗ (, )
- Shivani Verma, MD, MPH and
- Chukwudi Obiagwu, MBBS, MPH
- ↵∗Department of Business Management, Brooklyn College, 2900 Bedford Avenue, 218A, Brooklyn, New York 11210
The paper by Finkelman et al. (1) on cardiac dysfunction needs further clarification with regard to the statistical analyses used. First, analyses of the Wilcoxon rank sum test are used for comparisons between independent samples. However, Finkelman et al. (1) analyzed the same cohort over time. The correct analysis for within-sample analyses over time is the Wilcoxon signed-rank test. What would the p values in Table 3 be when the correct analysis of the Wilcoxon signed-rank test is used? Second, in their study limitations, Finkelman et al. (1) write that not all of their biomarkers were fully statistically normalized with their log-ratio transformation. They quote an article (2) that linear regression no longer needs to meet one of its basic assumptions of statistical normality. A reading of this article does not indicate that point. The article reviews what is commonly known; that is, that there can be some slight variations from normality assumptions for statistical analyses, such as allowing skewness values from 0 to 1 and kurtosis values ranging from –1 to 1, instead of only at the exact normality value of 0. The article conducted a simulation study to suggest that zero-inflated data (i.e., count data with many values of 0) can have variations from statistical normality when the data are leptokurtic. This is only 1 article and is not necessarily the standard in the field of statistics for zero-inflated data.
Also, there are questions that still need to be addressed for the metabolite data analyzed in the linear regression analyses in Finkelman et al. (1). What are the skewness and kurtosis values for the metabolites? Are the metabolites zero-inflated count data and leptokurtic? If the data have major variations from statistical normality and are not zero-inflated data and leptokurtic, the linear regression findings reported in Table 2 of Finkelman et al. are not meaningful according to all opinions.
Please note: The authors have reported that they have no relationships relevant to the contents of this paper to disclose.
- 2017 American College of Cardiology Foundation