abline and linearity over groups

?abline
and you can see
...
'reg' is a regression object which contains 'reg$coef'.  If it is
     of length 1 then the value is taken to be the slope of a line
     through the origin, otherwise, the first 2 values are taken to be
     the intercept and slope.
...

and
plot(test$l~test$t)
abline(lm(test$l~test$t))
(lm(test$l~test$t))
Call:
lm(formula = test$l ~ test$t)

Coefficients:
(Intercept)       test$t  
     0.4432     104.1688
test$tF=factor(test$t)
plot(test$l~test$tF)
abline(lm(test$l~test$tF))
(lm(test$l~test$tF))
Call:
lm(formula = test$l ~ test$tF)

Coefficients:
(Intercept)     test$tF1     test$tF2  
    -0.8776     108.1313     208.3376  

when test$tF is factor,these are 3 coef and the first two are used to drow the line,with  Intercept =   -0.8776 and slope= 108.1313   ,and     abline(lm(test$l~test$tF)) is abline(-0.8776,108.1313)

======= 2005-08-03 22:23:57 ÂÃ„ÂÃºÂÃ”ÂÃšÂÃ€ÂÂ´ÂÃÂÃ…ÂÃ–ÂÃÂÃÂÂ´ÂÂµÂÃ€ÂÂ£ÂÂº=======

Dear R users, please can you help me understand the behaviour of abline using function lm.

I'm trying to learn linearity over groups. So I make three groups with 10 values each:

test=data.frame(cbind(
l=c(rnorm(10,0,30),rnorm(10,100,30),rnorm(10,200,30)),
t = c(rep(0,10), rep(1,10), rep(2,10))
))

when I do: 

plot(test$l~test$t)
abline(lm(test$l~test$t))

the abline is a straight line through the centre of the points of each of the groups.

If, however, I factorise the groups (in order to do e.g. anova analysis) and then plot the data

test$tF=factor(test$t)
plot(test$l~test$tF)
abline(lm(test$l~test$tF))

the abline is now shifted up and to the left of where I would expect the line to go (through the centre of the points of each of the groups).

If there is a simple explanation, could someone tell me it?

---------------------------------
To help you stay safe and secure online, we've developed the all new Yahoo! Security Centre.
[[alternative HTML version deleted]]

______________________________________________
R-help at stat.math.ethz.ch mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html
= = = = = = = = = = = = = = = = = = = =

2005-08-03

------
Deparment of Sociology
Fudan University

Blog:http://sociology.yculblog.com
An embedded and charset-unspecified text was scrubbed...
Name: not available
Url: https://stat.ethz.ch/pipermail/r-help/attachments/20050803/ef99049d/attachment.pl
I think ronggui is right and your example is just a coincidence. Here is
my example in which case the intercept is hugely different 
(with slightly different style of coding).

  set.seed(1)  # for reproducibility 
  y  <- c( rnorm(10, 0, 30), rnorm(10, 100, 30), rnorm(10, 200, 30) )
  x  <- rep( 1:3, each=10 )

  df <- cbind.data.frame( y=y, x1=x, x2=factor(x) )
  plot(df$x2, df$y)
  points(df$x1, df$y, col=2, pch=2)

  ( fit1 <- lm( y ~ x1, data=df ) )
      (Intercept)           x1
           -89.55        96.01

  ( fit2 <- lm( y ~ x2, data=df ) )
      (Intercept)          x22          x23
            3.966      103.499      192.024

  abline(fit1)
  abline(fit2, col="red")  # wrong

The line above is wrong because it is fitting 

   abline(3.966, 103.499, col="green", lty=3)

as documented in help(abline) and pointed out by ronggui.

Note that 'fit1' is a linear model for regression while 
'fit2' is a linear model for ANOVA and that the documentation
of help(abline) uses the word "regression". Perhaps

It is more reliable to plot the fitted or predicted values via

   points( df$x2, fit2$fitted, col=4, pch=20 ) 

and this works regardless whether the linear model if for regression
or ANOVA.

You could replace plot() with lines() but this is perhaps not
appropriate with an ANOVA fit which may not have numerical values for x.

Regards, Adai
But those two lines are almose identical The difference between i=0.4432, s=104.1688 and i=0.8776,s=108.1313 is almost negligible.

What I see is that abline draws a line with a v.similar slope but intercept is about 90 instead of 0.8776. Try running the example to see what I mean.

ronggui <0034058 at fudan.edu.cn> wrote:>?abline
and you can see
...
'reg' is a regression object which contains 'reg$coef'. If it is
of length 1 then the value is taken to be the slope of a line
through the origin, otherwise, the first 2 values are taken to be
the intercept and slope.
...

and
plot(test$l~test$t)
abline(lm(test$l~test$t))
(lm(test$l~test$t))
Call:
lm(formula = test$l ~ test$t)

Coefficients:
(Intercept) test$t 
0.4432 104.1688 

test$tF=factor(test$t)
plot(test$l~test$tF)
abline(lm(test$l~test$tF))
(lm(test$l~test$tF))
Call:
lm(formula = test$l ~ test$tF)

Coefficients:
(Intercept) test$tF1 test$tF2 
-0.8776 108.1313 208.3376 

when test$tF is factor,these are 3 coef and the first two are used to drow the line,with Intercept = -0.8776 and slope= 108.1313 ,and abline(lm(test$l~test$tF)) is abline(-0.8776,108.1313)

======= 2005-08-03 22:23:57 ÂÃÂÂ´=======