Displaying 1 result from an estimated 1 matches for "numstab".
2019 Apr 12
2
integrate over an infinite region produces wrong results depending on scaling
...ng also failed, so I did both if
the first result was very small.
Today when integration of the rescaled function suddenly yielded a value
of 1.5 instead of 3.5 (not even zero). The MWE is below.
cons <- -0.020374721416129591
sc <- 0.00271245601724757383
sh <- 5.704
f <- function(x, numstab = 1) dgamma(cons - x * numstab, shape = sh,
scale = sc) * dgamma(-x * numstab, shape = sh, scale = sc) * numstab
curve(f, -0.06, 0, n = 501, main = "Unscaled f", bty = "n")
curve(f(x, sc), -0.06 / sc, 0, n = 501, main = "Scaled f", bty = "n")
sum(f(seq(-0....