[R] newton.method

[sammyny]

I have this function:
function(x)
-0.3*x*exp(-(((log(x)+(0.03+0.3*0.3/2)*0.5)/(0.3*sqrt(0.5)))^2)/2)/(2*sqrt(2*pi*0.5))
+ 0.03*exp(-0.03*0.5)*pnorm(-(log(x)+(0.03-0.3*0.3/2)*0.5)/(0.3*sqrt(0.5)))

uniroot is giving the correct results.
> uniroot(f,c(0,10))
$root
[1] 0.7347249
$f.root
[1] -1.955740e-07
$iter
[1] 12
$estim.prec
[1] 6.103516e-05

The newton method is not generating good result.
> for( x in seq(0,8,1)) { z <- nleqslv(x,f); print(c(z$x,z$fvec))} 
[1]           NaN 1.340781e+154
[1]           NaN 1.340781e+154
[1]  3.406704e+00 -5.608658e-09
[1]  3.340717e+00 -9.480243e-09
[1]  4.000000e+00 -5.470677e-11
[1]  5.000000e+00 -3.386274e-14
[1]  6.000000e+00 -3.559448e-17
[1]  7.000000e+00 -6.065006e-20
[1]  8.000000e+00 -1.581856e-22


In this case I am looking for largest value of 'x' such that f(x) >= 0

-- 
View this message in context: http://r.789695.n4.nabble.com/newton-method-tp2306111p2310041.html
Sent from the R help mailing list archive at Nabble.com.