[R] linear model with interaction / segments

[David Winsemius]

On Aug 29, 2009, at 7:56 AM, Markus Gesmann wrote:

> Dear R-help,
>
> Suppose I have the following data:
>
> df=data.frame(x=1:10, y=c(1,2,3,4,5,12,14,16,18,20))
> plot(y~x, df, t="b")
>
> How can I fit a model which estimates the slopes between x = 1-5,  
> 5-6, and 6-10?
>
> Adding the factor f:
>
> df$f <-  gl(2,5)
>
> Allows me to fit a linear model with interaction
>
> lm(y ~ x:f, data=df)
>
> which gives me the slope of 1 and 2 between 1-5, and 6-10  
> respectively, however it can not cope with the change from 5 to 6.

Can you be a bit more mathematically precise, ... and less  
anthropomorphic,  in explaining what you mean by "cannot cope"? When I  
look at the plot of fitted values I fail to see any "coping"  
deficiencies, at least with my understanding regarding what you were  
expecting .... about which I am very much guessing at this point. Did  
you want only two regression lines that share a joinpoint at x=5.5?

>
> I would appreciate if someone could point me into the right direction.
>
> Many thanks
>
> Markus

David Winsemius, MD
Heritage Laboratories
West Hartford, CT