Mercurial > hg > octave-thorsten
changeset 14817:8d3ab19f8599
Fix logncdf tests to use more accurate expected result.
* scripts/statistics/distributions/logncdf.m: use erf(1/sqrt(2)*1/sqrt(2))
instead of erf(1/2) to avoid floating point errors that are below eps
(alternate solution would be to add tolerance to the tests).
| author | Michael Goffioul <michael.goffioul@gmail.com> |
|---|---|
| date | Tue, 12 Jun 2012 19:50:13 +0100 |
| parents | 2feab289622a |
| children | 154c150f6a21 |
| files | scripts/statistics/distributions/logncdf.m |
| diffstat | 1 files changed, 1 insertions(+), 1 deletions(-) [+] |
line wrap: on
line diff
--- a/scripts/statistics/distributions/logncdf.m +++ b/scripts/statistics/distributions/logncdf.m @@ -73,7 +73,7 @@ %!shared x,y %! x = [-1 0 1 e Inf]; -%! y = [0, 0, 0.5, 1/2+1/2*erf(1/2), 1]; +%! y = [0, 0, 0.5, 1/2+1/2*erf((1/sqrt(2))*(1/sqrt(2))), 1]; %!assert (logncdf (x, zeros (1,5), sqrt(2)*ones (1,5)), y) %!assert (logncdf (x, 0, sqrt(2)*ones (1,5)), y) %!assert (logncdf (x, zeros (1,5), sqrt(2)), y)