On 30-08-2013, at 09:44, Jonsson <amen.alya...@bordeaux.inra.fr> wrote:
> I have three datasets that I want to compute the errors between them using
> linear regression.for this, I want to iterate to reach certain criteria for
> the calibration. if changes become smaller than eps the iteration is
> successful, hence stop and write parameters into cal:eps=0.00001 if number
> of iterations is > itermax the iteration failed, hence stop and fill cal
> with missing value itermax=400
>
> So I tried this code:
>
> x= c(5,2,4,2,1)
> y= c(5,3,4,6,9)
> z= c(5,8,4,7,3)
> itermax=400
> get initial calibration parameters, here we assume that:x is the reference
> dataset offset x_a=0, slope x_b=1 the other two datasets y, z are
> "calibrated" to x using a simple linear regression
>
> res=lm(x~y)
> y_a=coef(res)[1] ; y_b=coef(res)[2]
> res1=lm(x~z)
> z_a=coef(res1)[1] ; z_b=coef(res1)[2]
> y_t = y/y_b - y_a/y_b # "calibrate" y
> z_t = z/z_b - z_a/z_b #"calibrate" z
> x_e = sqrt(mean((x-y_t)*(x-z_t)))#calculate error of x
> iter <- 0
> while(((x_e-x) > 0.00001)&& (iter < itermax)) {
> iter <- 0 ##start iteration
> x = x_e
> res=lm(x~y)
> y_a=coef(res)[1] ; y_b=coef(res)[2]
> res1=lm(x~z)
> z_a=coef(res1)[1] ; z_b=coef(res1)[2]
> y_t = y/y_b - y_a/y_b # "calibrate" y
> z_t = z/z_b - z_a/z_b #"calibrate" z
> x_e = sqrt(mean((x-y_t)*(x-z_t)))
> iter <- iter + 1 # increase iteration counter
> }
> But I got the same result for X_e before and after the loop:
>
>> x_e
> [1] 6.454089
>
I tried your code and got this error message:
Error in model.frame.default(formula = x ~ y, drop.unused.levels = TRUE) :
variable lengths differ (found for 'y')
Calls: lm -> eval -> eval -> <Anonymous> -> model.frame.default
And looking at your code: x is a vector, x_e is a scalar and in the while loop
you are assigning x_e to x so x is then a scalar.
Berend
______________________________________________
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.
