UFYMA   27844
UNIDAD DE FITOPATOLOGIA Y MODELIZACION AGRICOLA
Unidad Ejecutora - UE
congresos y reuniones científicas
Título:
Non-linear mixed models implementation in InfoStat and interface to the nlme and lme4 libraries in R
Autor/es:
CASANOVES, F.; MACCHIAVELLI, R.; DI RIENZO, J.; BALZARINI, M.
Reunión:
Conferencia; XVII Conferencia Española y VII Encuentro Iberoamericano de Biometría; 2019
Resumen:
The nlme library of R implements linear and non-linear mixed models through the lme and nlme functions. The lme4 library implements linear, non-linear, and generalized linear mixed models using the lmer, nlmer, and glmer functions, respectively. For non-linear mixed effects models, the lme4 library uses Laplace and Gauss Hermite approximations to the likelihood, while the nlme library uses a pseudo-likelihood approach. Here we show the implementation, in the framework of InfoStat, of an interface to these functions to fit non- linear mixed effects models. When no random effects are present, the interface uses the nls function in a way that is transparent for the end user. Thus, this easy-to-use interface offers a complete non-linear modelling tool to InfoStat users. The implementation is complemented with a tutorial including several worked real life examples. The tutorial presents a short introduction to pseudo likelihood and likelihood estimation for non-linear mixed models, and discusses advantages and pitfalls of these methods using relevant examples. It also includes step-by-step instructions to use several options to obtain adjusted and predicted values as well as graphical tools for diagnostic purposes.