On Mon, Feb 16, 2009 at 5:28 PM, Patrick Giraudoux
<patrick.giraud...@univ-fcomte.fr> wrote:
> Greg Snow a écrit :
>> One approach is to create your own contrasts matrix:
>>
>>
>>> mycmat <- diag(8)
>>> mycmat[ row(mycmat) == col(mycmat) + 1 ] <- -1
>>> mycmati <- solve(mycmat)
>>> contrasts(agefactor) <- mycmati[,-1]
>>>
>>
>> Now when you use agefactor, the intercept will be the first age group and
>> the slopes will be the differences between the pairs of groups (make sure
>> that the order of the levels of agefactor is correct).
>>
>> The difference between this method and the contr.sdif function in MASS is
>> how the intercept will end up being interpreted (and the dimnames).
>>
>> Hope this helps,
>>
>>
>
> Actually, exactly what I needed including the reference to contr.sdif in
> MASS I did not spot before (although I am a faithful reader of the
> yellow book... but so many things still escape to me). Again thanks a lot.
>
> Patrick
>
Glad you were able to solve your problem. Frank replied earlier with
the suggestion to use contrast.Design() to perform the tests after the
initial model has been fit. Perhaps I am a little denser than normal,
but I cannot figure out how to apply contrast.Design() to a similar
model with several levels of a factor. Example data:
# need these
library(lattice)
library(Design)
# replicate an important experimental dataset
set.seed(10101010)
x <- rnorm(100)
y1 <- x[1:25] * 2 + rnorm(25, mean=1)
y2 <- x[26:50] * 2.6 + rnorm(25, mean=1.5)
y3 <- x[51:75] * 2.9 + rnorm(25, mean=5)
y4 <- x[76:100] * 3.5 + rnorm(25, mean=5.5)
d <- data.frame(x=x, y=c(y1,y2,y3,y4), f=factor(rep(letters[1:4], each=25)))
# plot
xyplot(y ~ x, groups=f, data=d, auto.key=list(columns=4, title='Beard
Type', lines=TRUE, points=FALSE, cex=0.75), type=c('p','r'),
ylab='Number of Pirates', xlab='Distance from Land')
# standard comparison to base level of f
summary(lm(y ~ x * f, data=d))
# compare to level 4 of f
summary(lm(y ~ x * C(f, base=4), data=d))
# now try with contrast.Design():
dd <- datadist(d)
options(datadist='dd')
l <- ols(y ~ x * f, data=d)
# probably the wrong approach, as the results do not look too familiar:
contrast(l, list(f=levels(d$f)))
x f Contrast S.E. Lower Upper t Pr(>|t|)
-0.3449623 a 0.3966682 0.1961059 0.007184856 0.7861515 2.02 0.0460
-0.3449623 b 0.5587395 0.1889173 0.183533422 0.9339456 2.96 0.0039
-0.3449623 c 4.1511677 0.1889170 3.775962254 4.5263730 21.97 0.0000
-0.3449623 d 4.3510407 0.1888820 3.975904726 4.7261766 23.04 0.0000
This is adjusting the output to a given value of 'x'... Am I trying to
use contrast.Design() for something that it was not intended for? Any
tips Frank?
Cheers,
Dylan
______________________________________________
R-help@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.
