and finally...
thingy <- function(x) {
x <- C(x, poly, 1)
tmp <- contrasts(x)
contrasts(x, 1) <- 2 * tmp / sum(abs(tmp))
x
}
dat2 <- with(data.catapult,
data.frame(
Distance,
h=thingy(h),
s=thingy(s),
l=thingy(l),
e=thingy(e),
b=thingy(b)
)
)
lm5 <- lm(Distance ~ .^2, data = dat2)
summary(lm5)
On Tue, Jun 26, 2012 at 12:35 AM, Simon Knapp <sleepingw...@gmail.com> wrote:
> ... but this is tantalisingly close:
>
> dat1 <- with(data.catapult,
> data.frame(
> Distance,
> h=C(h, poly, 1),
> s=C(s, poly, 1),
> l=C(l, poly, 1),
> e=C(e, poly, 1),
> b=C(b, poly, 1)
> )
> )
> lm4 <- lm(Distance ~ .^2, data = dat1)
> summary(lm4)
>
> ... wish I knew what it meant.
>
>
>
> On Tue, Jun 26, 2012 at 12:18 AM, Simon Knapp <sleepingw...@gmail.com> wrote:
>> They are coding the variables as factors and using orthogonal
>> polynomial contrasts. This:
>>
>> data.catapult <- data.frame(data.catapult$Distance,
>> do.call(data.frame, lapply(data.catapult[-1], factor, ordered=T)))
>> contrasts(data.catapult$h) <-
>> contrasts(data.catapult$s) <-
>> contrasts(data.catapult$l) <-
>> contrasts(data.catapult$e) <-
>> contr.poly(3, contrasts=F)
>> contrasts(data.catapult$b) <- contr.poly(2, contrasts=F)
>> lm1 <- lm(Distance ~ .^2, data=data.catapult)
>> summary(lm1)
>>
>> gets you closer (same intercept at least), but I can't explain the
>> remaining differences. I'm not even sure why the results to look like
>> they do (interaction terms like "a*b" not "a:b" and one level for each
>> interaction).
>>
>> Hope that helps,
>> Simon Knapp
______________________________________________
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.
