[R] exponential model in R

[Ben Bolker]

  Please continue the conversation on-list.

  (1) Various things can happen, but in particular you need to identify
good starting parameters. 'a' should be approximately the x-intercept, b
should be approximately the slope of log(proc) as cls changes.

 (2) I don't know what "impossible to put it in the plot with lines
command" means.

 (3) Please read the posting guide and this URL:
http://tinyurl.com/reproducible-000 about asking reproducible questions.
 You should also try googling "r nonlinear regression examples" and read
some of the material you find ...


On 13-02-10 01:51 PM, catalin roibu wrote:
> Dear Ben Bolker,
> 
> I don't know what is happening wen I run this script: n1 <-
> nls(proc~a*exp(b*cls),start=list(a=4,b=-0.01),data=dpoz)
> Error in nls(proc ~ a * exp(b * cls), start = list(a = 4, b = -0.01),  : 
>   singular gradient
> 
> Please help me to solve this problem!
> 
> Thank you!
> 
> 
> On 10 February 2013 18:17, catalin roibu <catalinroibu at gmail.com
> <mailto:catalinroibu at gmail.com>> wrote:
> 
>     Dear Ben,
>     Thank you for your help!
>     I have two problems if I run the script. If I use the first equation
>     I have this error: Error in nls(proc ~ a * exp(b * cls), start =
>     list(a = 4, b = -0.01),  : 
>       singular gradient
> 
>     If I use the second equation is OK but is impossible to put it in
>     the plot with lines command.
> 
>     Thank you!
> 
> 
>     On 10 February 2013 17:10, Ben Bolker <bbolker at gmail.com
>     <mailto:bbolker at gmail.com>> wrote:
> 
>         catalin roibu <catalinroibu <at> gmail.com <http://gmail.com>>
>         writes:
> 
>         >
>         > Dear R users,
>         > I don't know how to compute an exponential model like this:
>         > proc=a*exp(b*cls), or proc=a*exp(b*cls)+c*exp(d*cls). Please
>         help me to
>         > solve this problem!
>         >
>         > Thank you!
>         >
>         > My data is:
> 
>           This still doesn't make much sense/have much context,
>         but I'll make some guesses.
> 
>         dat <- read.table(textConnection("
>         proc cls
>         0.5 452.616206 0.5
>         1 255.864021 1.0
>         1.5 150.885316 1.5 <tel:150.885316%201.5>
>         2 86.289600 2.0 <tel:86.289600%202.0>
>         2.5 56.321559 2.5 <tel:56.321559%202.5>
>         3 39.504444 3.0
>         3.5 25.570308 <tel:3.5%2025.570308> 3.5
>         4 5.382726 4.0
>         "),header=TRUE)
> 
>         with(dat,plot(cls~proc))
> 
>         I suspect that what you want to *fit* these models.
> 
>         n1 <- nls(cls~a*exp(b*proc),start=list(a=4,b=-0.01),data=dat)
>         n2 <- nls(cls~SSbiexp(proc,a,logb,c,logd),data=dat)
> 
>         lines(dat$proc,predict(n1))
>         lines(dat$proc,predict(n2),col=2)
> 
>         ______________________________________________
>         R-help at r-project.org <mailto:R-help at 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.
> 
> 
> 
> 
>     -- 
>     ---
>     Catalin-Constantin ROIBU
>     Forestry engineer, PhD
>     Forestry Faculty of Suceava
>     Str. Universitatii no. 13, Suceava, 720229, Romania
>     office phone     +4 0230 52 29 78, ext. 531
>     <tel:%2B4%200230%2052%2029%2078%2C%20ext.%20531>
>     mobile phone   +4 0745 53 18 01 <tel:%2B4%200745%2053%2018%2001>
>                            +4 0766 71 76 58 <tel:%2B4%200766%2071%2076%2058>
>     FAX:                +4 0230 52 16 64 <tel:%2B4%200230%2052%2016%2064>
>     silvic.usv.ro <http://silvic.usv.ro>
> 
> 
> 
> 
> -- 
> ---
> Catalin-Constantin ROIBU
> Forestry engineer, PhD
> Forestry Faculty of Suceava
> Str. Universitatii no. 13, Suceava, 720229, Romania
> office phone     +4 0230 52 29 78, ext. 531
> mobile phone   +4 0745 53 18 01
>                        +4 0766 71 76 58
> FAX:                +4 0230 52 16 64
> silvic.usv.ro <http://silvic.usv.ro>