Fig. 1. Relationship between twinning status and the number of maternal total births.
The plots represent marginal predictions (a: points, b: line) ± CI95% (a: error bars, b: grey area) from the fits of generalised linear mixed-effects models including the variable depicted on the x-axis as fixed effect and variation between populations as a random effect (n = 21,290 mothers in total, from 8 populations; see Table 1). a Twinners had more total births (6.29; CI95%: 5.89, 6.72) than non-twinners (4.86; CI95%: 4.58, 5.12), and (b) the probability of a mother being a twinner was positively related to a mother’s total number of births. Note that the y-axis is represented on a logit scale so as to display the outcome of the logistic (binary) regression as a straight line with an estimated slope β of 0.162 (CI95%: 0.145, 0.178). Model summary statistics are provided in Supplementary Tables 1, 2.