Have you tried Eureqa?
http://ccsl.mae.cornell.edu/eureqa
It tries to discover the model from your data.
Try it, it is impressive (at least on my data).
Ciao!
mario
On 14-May-10 15:07, Thomas Levine wrote:
Actually, ignore my comment about that link. I don't think that link
is wh
Actually, ignore my comment about that link. I don't think that link
is what you want to look at either.
y~x^2 fits quite well, but you could also write a loop to run lm() on
a bunch of different transformations.
foo=list(log,sqrt)
for (bar in foo) {
plot(bar(x),y)
}
There may be a funct
It appears that as one proceeds from right to left that it flattens
out at 0.021 so lets try this where we have added a bit to 0.021 to
avoid log(0)
plot(x, log(0.0211 - y))
and
plot(1/x, log(0.0211 - y))
Except for the first point the latter plot looks linear so lets try:
fm <- nls(y ~ cbind(
I spoke too soon; that is not linear.
I've never used the model-fitting functions, but you may check them out.
http://developer.r-project.org/model-fitting-functions.txt
Tom
2010/5/14 Thomas Levine :
> x<-c(0.5,4,6,8,12)
> y<-c(0.021,0.021,0.020,0.018,0.012)
> lm(y~x)
>
> 2010/5/14 Dani Valverde
x<-c(0.5,4,6,8,12)
y<-c(0.021,0.021,0.020,0.018,0.012)
lm(y~x)
2010/5/14 Dani Valverde :
> Hello,
> It is a very naive question, but here it is. I have this values:
>
> x: 0.5 4 6 8 12
>
> y: 0.021 0.021 0.020 0.018 0.012
>
> I need to fit a function to them. How can I do it with R?
> Thank you so
5 matches
Mail list logo
