README.md

In lian0090/FW: Performs Gibbs Sampler and Least Square models for Finlay-Wilkinson regressions

FW

Bayesian method and ordinary least square method for Finlay-Wilkinson Regression.

Install

library(devtools)
install_github("lian0090/FW")

Note, install_github by default does not build vignettes. To be able to directly view vignettes from R, you need to do the following installation instead. install_github("lian0090/FW",build_vignettes=T)

Basic usage

library(FW)
data(wheat)
attach(wheat.Y)
lm1=FW(y=y,VAR=VAR,ENV=ENV)
plot(lm1)

FAQ

Why the estimate of genetic effect g from Drdinary Least Squares much larger than from Bayesian method?

The FW regression was fitting the model y=mu+g+(b+1)h+e

In Ordianary Least Square method, a linear regression model is fitted within each line/variety to estimate the main genetic effect g where mu is set to 0 for this within line/variety linear regression and g is the intercept of each within line regression. However, in the Bayesian method, the whole data set was used to fit the model, and the g is estimated as a random effect with zero mean. Therefore, the estimated g from ordinary linear regression is generally much larger than the g estimated from Bayesian method due to fact that the estimated g from ordinary linear regression contains an overall mean.

Detailed implementation can be found in the vignettes of the package.

lian0090/FW documentation built on May 20, 2019, 5:27 p.m.