Premium
Nonlinear Modeling for Analyzing Data from Multiple Harvest Crops
Author(s) -
Sari Bruno G.,
Olivoto Tiago,
Diel Maria I.,
Krysczun Dionatan K.,
Lúcio Alessandro D. C.,
Savian Taciana V.
Publication year - 2018
Publication title -
agronomy journal
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.752
H-Index - 131
eISSN - 1435-0645
pISSN - 0002-1962
DOI - 10.2134/agronj2018.05.0307
Subject(s) - gompertz function , nonlinear system , statistics , mathematics , inflection point , parametric model , parametric statistics , production (economics) , sigmoid function , goodness of fit , logistic function , econometrics , computer science , physics , geometry , quantum mechanics , machine learning , artificial neural network , economics , macroeconomics
We proposed a workflow for nonlinear modeling of data from multiple‐harvest crops. We demonstrated why the nonlinearity measures should be used to select nonlinear models. We demonstrated as the critical points describe the multiple‐harvest crops production. Logistic model parameters determine the precocity and the concentration of production. Growth models are alternative to ANOVA in analyzing data from multiple‐harvest crops.Nonlinear growth models have been widely used for analyzing production curves with a sigmoidal pattern; however, all benefits that these models provide are not being fully exploited. Our aim here is to provide a step‐by‐step guide on how to choose a nonlinear model with parameters close to being unbiased, and to show how to estimate and interpret the critical points of a model aimed at determining the precocity and concentration of the production. Data on two uniformity trials conducted with eggplant ( Solanum melongena L.) was used for this purpose. The Brody, Gompertz, logistic, and von Bertalanffy models were fitted to predict the number and fresh mass of fruits per plant. The model with lower nonlinearity measures and lower bias of the parameter estimates was selected. All the tested models presented satisfactory goodness‐of‐fit measures, but they differed regarding nonlinearity measures. The logistic model was selected because it had lower intrinsic and parametric nonlinearity and lower bias in parameter estimates. The inflection point and maximum acceleration/deceleration points of this model provide detailed pieces of information of the production through the productive cycle. Finally, using the logistic model as an example, we demonstrate that lower values of β 2 are related to an earlier maximum production rate, and higher values of β 3 are related to an earlier production that is concentrated in fewer days. The nonlinearity measures were important for the model selection. Thus, it is strongly recommended that nonlinearity is estimated and used to select nonlinear models in future studies.