A comparative study between nonlinear regression and nonparametric approaches for modelling Phalaris paradoxa seedling emergence

Parametric nonlinear regression (PNR) models are used widely to fit weed seedling emergence patterns to soil microclimatic indices. However, such approximation has been questioned, mainly due to several statistical limitations. Alternatively, nonparametric approaches can be used to overcome the problems presented by PNR models. Here, we used an emergence data set of Phalaris paradoxa to compare both approaches. Mean squared error and correlation results indicated higher accuracy for the descriptive ability but similar poor performance for predictive ability of the nonparametric approach in comparison with the PNR approach. These results suggest that our nonparametric cumulative distribution function approach is a valuable alternative to the classical parametric nonlinear regression models to describe complex emergence patterns for P. paradoxa, but not to predict them.