metan (version 1.18.0)

predict.waas: Predict the means of a waas object

Description

Predict the means of a waas object considering a specific number of axis.

Usage

# S3 method for waas
predict(object, naxis = 2, ...)

Value

A list where each element is the predicted values by the AMMI model for each variable.

Arguments

object

An object of class waas

naxis

The the number of axis to be use in the prediction. If object has more than one variable, then naxis must be a vector.

...

Additional parameter for the function

Author

Tiago Olivoto

Details

This function is used to predict the response variable of a two-way table (for examples the yielding of the i-th genotype in the j-th environment) based on AMMI model. This prediction is based on the number of multiplicative terms used. If naxis = 0, only the main effects (AMMI0) are used. In this case, the predicted mean will be the predicted value from OLS estimation. If naxis = 1 the AMMI1 (with one multiplicative term) is used for predicting the response variable. If naxis = min(gen-1;env-1), the AMMIF is fitted and the predicted value will be the cell mean, i.e. the mean of R-replicates of the i-th genotype in the j-th environment. The number of axis to be used must be carefully chosen. Procedures based on Postdictive success (such as Gollobs's d.f.) or Predictive sucess (such as cross-validation) should be used to do this. This package provide both. waas() function compute traditional AMMI analysis showing the number of significant axis. On the other hand, cv_ammif() function provide a cross-validation, estimating the RMSPD of all AMMI-family models, based on resampling procedures.

Examples

# \donttest{
library(metan)
model <- waas(data_ge,
             env = ENV,
             gen = GEN,
             rep = REP,
             resp = c(GY, HM))
# Predict GY with 3 IPCA and HM with 1 IPCA
predict <- predict(model, naxis = c(3, 1))
predict
# }