Fig. 2: High-ranked genera are more stable with increasing mean abundance than low-ranked genera.
Mean-variance relationships can be characterized by a linear relationship known as Taylor’s Law (TL). According to TL log(variance)g = b × log(mean)g, where b is a species-specific constant and g indicates the group of measurements. For a given unit change in the log(mean) of population abundance, the TL’s slope equals the change in log(variance), which measures heterogeneity or scatter in the distribution of population abundance. A greater slope in the temporal TL means a greater degree of change in the temporal variance of genus population abundance with respect to its temporal mean. Genera for which the TL slope is 2 have about the same coefficient of variation in all subjects. Genera for which the TL slope is less than 2 show more temporal variation in subjects in which they are more abundant. The reverse for genera for which the TL slope is greater than 2. Here, we show Taylor slopes of high ranked genera (over all individuals) are generally higher, indicating more variation with increasing mean abundance, compared to low ranked genera. a Mean-variance relationships for all genera satisfying the abundance and prevalence criteria (genera with abundance >0.5%, present in >5 individuals in >3 samples). Regression lines are colored according to Taylor’s slope from less steep (blue) to steep (red). b TL slope of high versus low-ranked genera, for the top 10, 20, and 50 genera (t-test between TL slopes of top 10, 20, and 50 versus lower-ranked genera, two-sided, P = 0.00318, P = 0.0000285, P = 0.000188, respectively, N = 73). c Intra-individual variation in alpha diversity measurements, (top) observed richness with an indication of the total number of genera detected over all samples (squares), (middle) Pielou’s evenness index ‘J’, and (below) Shannon alpha diversity index for each individual (ID-number 800–821) (N = 20, n = 694). The body of the box plots represents the first and third quartiles of the distribution, and the median line. Whiskers extend from the quartiles to the last data point within 1.5×IQR, with outliers beyond. Significance levels: ***: 0.001, **: 0.01, *: 0.05.
