Association between Relative Amounts of White Blood Cell Counts: a Case of Distribution Dependent Correlations

Abstract

We previously observed a positive association between relative amounts of particular body fatty acids, provided they had low-numbers relative to sum of the remaining ones. Additionally, theoretical considerations and computer experiments suggested that, with two high-number variables relative to one with much lower numbers, we should expect a negative relationship between relative amounts of the high-number variables. Moreover, the correlation outcomes with true values could be well reproduced using random numbers, provided that the numbers had the true ranges (distributions). This finding led to the concept of Distribution Dependent Correlations. Since counts of segmented neutrophil leukocytes (N) and lymphocytes (L) are normally much higher than sum of the remaining (R) white blood cells (WBC), it was suggested that %N might possibly be negatively associated with %L. In the present work, random numbers were sampled in lieu of reported WBC subgroup values, but using the measured mean ± SD values. The results demonstrate that relative amounts of random number “N” and “L” were indeed inversely related in both sexes: Spearman’s rho = -0.9, p <0.001, n = 200, as observed using within-person data, and between-person data as well. Furthermore, an alteration in distributions (variability) of the WBC subgroup changed the correlation outcome, as evaluated by scatterplots and correlation coefficients. Decreasing (increasing) values of %R improved (made poorer) the negative association between %N and %L. Thus, the observed negative association between %N and %L seems to be a case of Distribution Dependent Correlations. Hypothetically, by directing WBC subgroup counts to particular places on the scale, a powerful tool is available to govern the associations between relative amounts of WBC subgroups.