🤖 AI Summary
Existing R tools lack systematic support for crossover design data—particularly those involving longitudinal within-period measurements and carryover effects. This paper introduces CrossCarry, the first open-source R package specifically designed for modeling crossover trials. It accommodates exponential-family responses, arbitrary-order designs, and scenarios with or without washout periods. Methodologically: (1) it extends the generalized estimating equations (GEE) framework by jointly modeling within-period correlation structures and between-period carryover dependencies—a novel integration; (2) it incorporates B-spline–based nonparametric components to flexibly estimate both temporal trends and carryover effects; and (3) it enables unified, flexible modeling of treatment, time, and carryover effects. Empirical evaluations demonstrate that CrossCarry substantially improves statistical power and estimation accuracy for treatment and carryover effects under challenging conditions—including skewed responses and weak washout.
📝 Abstract
Experimental crossover designs are widely used in medicine, agriculture, and other areas of the biological sciences. Due to the characteristics of the crossover design, each experimental unit has longitudinal observations and the presence of drag effects on the response variable. There is no package in {R} that clearly models data from crossover designs. The {CrossCarry} package presented in this paper allows testing any crossover design as long as the observed response variable belongs to the exponential family, regardless of whether or not there is a washout period. It also allows modeling repeated measurements within each period and extends the correlation structures used in the generalized estimating equations. The family of correlation structures is built that takes into account the particularities of the design, that is, the correlation between and within the periods. It also includes a parametric component for modeling treatment effects and a non-parametric component for modeling time effects and carry-over effects. The non-parametric component is estimated from splines inserted into the generalized estimation equations.