Using regression trees to predict patterns of male provisioning in the stitchbird (hihi)

Abstract

Males are predicted to trade-off parental effort (PE) for current or future mating effort (ME) at times when this is likely to increase their lifetime reproductive success. We used a cross-validated regression tree analysis to predict the effect of several variables on nest-provisioning visits by male stitchbirds (or hihi, Notiomystis cincta). For 51 first broods, males did not invest in secondary females' broods, nor did they provision nests containing only one chick. Male PE was positively correlated with brood size in nests containing more than one chick: male age, additional mating opportunities and proximity to supplementary food had minor effects on male provisioning rates in moderate-sized broods. For 23 second broods, provisioning was also correlated with female rank and brood size. Regression trees offer advantages over traditional linear parametric techniques. In this study, they outperformed multiple linear regression in their prediction accuracy, in highlighting interactions and allowing visualization of a hierarchy of decision-making rules used to predict offspring provisioning by male stitchbirds. While previous models of paternal investment have focused on ME/PE trade-offs as static functions, we analysed first and second broods separately because of the possibility that the relative importance of some variables changes as the breeding season progresses.