Thanks for the pointer - the 'lm(y~x:z)' model does give the slopes directly and hence confint gives the confidence intervals.
The thing that puzzles me is that my dummy data explicitly sets the three levels of the factor to have different variances and yet the standard error is the same for all three parameter estimates in the summary.lm output - is this a common standard error of the 'x:z' term in the model? If you fit a separate regression to subsets of the data for each level in 'z' then the standard errors of the slope reflect these differences in variance. What I was trying to get was confidence limits from within a single model that also reflect the difference in certainty about the three slopes.
I realize that this is a failing of my understanding and more a stats question than an R question - if anyone can give me any advice or pointers, that would be great.
Thanks, David
On 29 Mar 2004, at 20:04, BXC (Bendix Carstensen) wrote:
You may want:
lm( y ~ x:z )
This is the same model you fitted, but prametrized differently. But please check that what you REALLY want is not
lm( y ~ z + x:z )
This is the model with different intercepts as well.
Bendix Carstensen ---------------------- Bendix Carstensen Senior Statistician Steno Diabetes Center Niels Steensens Vej 2 DK-2820 Gentofte Denmark tel: +45 44 43 87 38 mob: +45 30 75 87 38 fax: +45 44 43 07 06 [EMAIL PROTECTED] www.biostat.ku.dk/~bxc ----------------------
-----Original Message----- From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On Behalf Of David Orme Sent: Monday, March 29, 2004 4:44 PM To: [EMAIL PROTECTED] Subject: [R] Confidence Intervals for slopes
Hi,
I'm trying to get confidence intervals to slopes from a linear model and I can't figure out how to get at them. As a cut 'n' paste example:
################# # dummy dataset - regression data for 3 treatments, each treatment with different (normal) variance x <- rep(1:10, length=30) y <- 10 - (rep(c(0.2,0.5,0.8), each=10)*x)+c(rnorm(10, sd=0.1), rnorm(10, sd=0.6),rnorm(10, sd=1.1)) z <- gl(3,10) plot(y~x, pch=unclass(z))
# model as three slopes with common intercept options(contrasts=c("contr.treatment","contr.poly")) model <- lm(y~x+x:z)
# coefficient table in summary gives the intercept, first slope and the difference in slopes summary(model)
# confint gives the confidence interval for the intercept and first slope, # and the CIs for the _differences_ confint(model) #################
What I'd like to report are the actual CI's for the slopes for the second and third treatments, in the same way that confint returns the parameter estimates for the first treatment. Can anyone point me in the right direction?
Thanks, David
______________________________________________ [EMAIL PROTECTED] mailing list https://www.stat.math.ethz.ch/mailman/listinfo> /r-help PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html
______________________________________________ [EMAIL PROTECTED] mailing list https://www.stat.math.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html
