Mark Klose | Search arXiv

arXiv2026

Delicatessen: Automated Estimating Equations in Python

Estimating equation theory provides a unified framework for statistical modeling and inference. Applying estimating equations has historically involved extensive matrix algebra and by-hand differentiation of complex functions. Here, we introduce delicatessen, a Python library that automates those tedious calculations, lowering the barrier to adoption in quantitative biological and life science research. To highlight the utility of delicatessen for the acceleration of scientific research, we provide illustrative examples of linear regression with outliers, estimation of a dose-response curve, and standardization of the mean. Across these and other settings, delicatessen streamlines transparent and reproducible advanced statistical procedures for modern data analysis.

stat.ME↗

arXiv2024

Healthy Live Births Should be Considered as Competing Events when Estimating the Total Effect of Prenatal Medication Use on Pregnancy Outcomes

Pregnancy loss is recognized as an important competing event in studies of prenatal medication use. However, a healthy live birth also precludes subsequent adverse pregnancy outcomes, yet these events are often censored. Using Monte Carlo simulation, we examine bias that results from failure to account for healthy live birth as a competing event in estimates of the total effect of prenatal medication use on pregnancy outcomes. We simulated data for 12 trials estimating the effect of antihypertensive initiation versus non-initiation on two outcomes: (1) composite fetal death or severe prenatal preeclampsia and (2) small-for-gestational-age (SGA) live birth. We used time-to-event methods to estimate absolute risks, risk differences and risk ratios. For the composite outcome, we conducted two analyses where non-preeclamptic live birth was (1) a censoring event and (2) a competing event. For SGA live birth, we conducted three analyses where fetal death and non-SGA live birth were (1) censoring events, (2) a competing event and censoring event, respectively; and (3) competing events. In all analyses, censoring healthy live births led to inflated absolute risk estimates as well as bias and imprecise treatment effect estimates. Studies of prenatal exposures on pregnancy outcomes should analyze healthy live births as competing risks to estimate unbiased total treatment effects.

stat.AP↗