Forecasting and predictive inference are fundamental data analysis tasks. Most studies employ parametricapproaches making strong assumptions about the data generating process. On the other hand, while nonparametric modelsare applied, it is sometimes found in situations involving low signal to noise ratios or large numbers of covariates that theirperformance is unsatisfactory. We propose a new varying-coefficient semiparametric model averaging prediction (VC-SMAP)approach to analyze large data sets with abundant covariates. Performance of the procedure is investigated with numericalexamples. Even though model averaging has been extensively investigated in the literature, very few authors have consideredaveraging a set of semiparametric models. Our proposed model averaging approach provides more flexibility than parametricmethods, while being more stable and easily implemented than fully multivariate nonparametric varying-coefficient models.We supply numerical evidence to justify the effectiveness of our methodology.

Thresholding variable plays a crucial role in subgroup identification for personalizedmedicine. Most existing partitioning methods split the sample basedon one predictor variable. In this paper, we consider setting the splitting rulefrom a combination of multivariate predictors, such as the latent factors, principlecomponents, and weighted sum of predictors. Such a subgrouping methodmay lead to more meaningful partitioning of the population than using a singlevariable. In addition, our method is based on a change point regression modeland thus yields straight forward model-based prediction results. After choosinga particular thresholding variable form, we apply a two-stage multiple changepoint detection method to determine the subgroups and estimate the regressionparameters. We show that our approach can produce two or more subgroupsfrom the multiple change points and identify the true grouping with high probability.In addition, our estimation results enjoy oracle properties. We design asimulation study to compare performances of our proposed and existing methodsand apply them to analyze data sets from a Scleroderma trial and a breastcancer study.

Background

Socioeconomic disparities in infant mortality have persisted for decades in high-income countries and may have become stronger in some populations. Therefore, new understandings of the mechanisms that underlie socioeconomic differences in infant deaths are essential for creating and implementing health initiatives to reduce these deaths. We aimed to explore whether and the extent to which preterm birth (PTB) and small for gestational age (SGA) at birth mediate the association between maternal education and infant mortality.

Methods and findings

We developed a population-based cohort study to include all 1,994,618 live singletons born in Denmark in 1981-2015. Infants were followed from birth until death, emigration, or the day before the first birthday, whichever came first. Maternal education at childbirth was categorized as low, medium, or high. An inverse probability weighting of marginal structural models was used to estimate the controlled direct effect (CDE) of maternal education on offspring infant mortality, further split into neonatal (0-27 days) and postneonatal (28-364 days) deaths, and portion eliminated (PE) by eliminating mediation by PTB and SGA. The proportion eliminated by eliminating mediation by PTB and SGA was reported if the mortality rate ratios (MRRs) of CDE and PE were in the same direction. The MRRs between maternal education and infant mortality were 1.63 (95% CI 1.48-1.80, P < 0.001) and 1.19 (95% CI 1.08-1.31, P < 0.001) for low and medium versus high education, respectively. The estimated proportions of these total associations eliminated by reducing PTB and SGA together were 55% (MRRPE = 1.27, 95% CI 1.15-1.40, P < 0.001) for low and 60% (MRRPE = 1.11, 95% CI 1.01-1.22, P = 0.037) for medium versus high education. The proportions eliminated by eliminating PTB and SGA separately were, respectively, 46% and 11% for low education (versus high education) and 48% and 13% for medium education (versus high education). PTB and SGA together contributed more to the association of maternal educational disparities with neonatal mortality (proportion eliminated: 75%-81%) than with postneonatal mortality (proportion eliminated: 21%-23%). Limitations of the study include the untestable assumption of no unmeasured confounders for the causal mediation analysis, and the limited generalizability of the findings to other countries with varying disparities in access and quality of perinatal healthcare.

Conclusions

PTB and SGA may play substantial roles in the relationship between low maternal education and infant mortality, especially for neonatal mortality. The mediating role of PTB appeared to be much stronger than that of SGA. Public health strategies aimed at reducing neonatal mortality in high-income countries may need to address socially related prenatal risk factors of PTB and impaired fetal growth. The substantial association of maternal education with postneonatal mortality not accounted for by PTB or SGA could reflect unaddressed educational disparities in infant care or other factors.