gls does not give correct intercept and slope during multiple regression

Asked Aug 26 '17 at 00:23

Active Aug 26 '17 at 00:23

Viewed 166 times

Anybody could please tell me why gls function sometimes does not give the correct intercept and slope for a given dataset? Here is a reproducible example:

  require(nlme)
  s=read.csv("exampleglsabline.csv")
  m1=gls(log(y)~log(x2)+log(x1),method="REML",data =  s)
  summary(m1)
  x=s$x1
  y=residuals(gls(log(y)~log(x2),method="REML",data =  s))
  plot(x,y,xlab="x1",ylab=expression("y"),
  type = 'p',bg=adjustcolor("seashell", alpha=0.8), col = 'black',cex=2, pch=21,cex.lab=1,
  xlim=c(min(0),max(x)),
  ylim=c(min(y),max(y)))
  abline(a=7.251864,b=0.882575)

asked Aug 26 '17 at 00:23

Agus camacho

1

scale your variables: `scale(log(y))~scale(log(x2))`, and have `x=log(s$x1)` – user2957945 Aug 26 '17 at 01:04
First get the example straight. Then provide the data so that we can reproduce your result. – Roman Luštrik Aug 26 '17 at 06:14
Thanks user2957945, this works indeed. Would you tell me why the scaling is necessary? and why only scaling the data , without taking logs first, leads to different relationships? – Agus camacho Aug 26 '17 at 14:33
@ Roman Luštrik, I did not understand your answer. The data was available and the example runs smoothly with it. – Agus camacho Aug 27 '17 at 14:02

gls does not give correct intercept and slope during multiple regression

0 Answers0