[R] interpolation to abscissa

[e-letter]

On 13/01/2009, David Winsemius <dwinsemius at comcast.net> wrote:
> It's fairly clear from the documentation that approxfun() will not
> extrapolate.
>
> help.search("extrapolate")
> library(Hmisc)
> ?approxExtrap
>
> Some sort of minimization approach:
>
>  > approxExtrap(x=c(0,5,10,15,20), y=c(16,45,77,101,125),xout=c(-4,0,4))
> $x
> [1] -4  0  4
>
> $y
> [1] -7.2 16.0 39.2
>
>  > approxExtrap(x=c(0,5,10,15,20),
> y=c(16,45,77,101,125),xout=seq(-2.8,-2.6, by=0.01))
> $x
>   [1] -2.80 -2.79 -2.78 -2.77 -2.76 -2.75 -2.74 -2.73 -2.72 -2.71
> -2.70 -2.69 -2.68
> [14] -2.67 -2.66 -2.65 -2.64 -2.63 -2.62 -2.61 -2.60
>
> $y
>   [1] -0.240 -0.182 -0.124 -0.066 -0.008  0.050  0.108  0.166  0.224
> 0.282  0.340
> [12]  0.398  0.456  0.514  0.572  0.630  0.688  0.746  0.804  0.862
> 0.920
>
> How accurate do you need the answer?
>
> I tried Hmisc's inverseFunction(), but it returned 0 for an argument
> of zero:
>
>  > invF <- inverseFunction(x=c(0,5,10,15,20), y=c(16,45,77,101,125))
>  > invF(0)
>
> So I then hacked Harrell's inverseFunction by substituting
> approxExtrap in every in instance
> where approx appeared, creating invFunc2:
>
> then
>
>  > invF <- invFunc2(x=c(0,5,10,15,20), y=c(16,45,77,101,125))
>  >
>  > invF(0)
> [1] -2.758621
I have compared your answer to those obtained from gnuplot, scilab and
qtiplot; all report a result of x=-3.28. Why is r different?