Skip to content
Prev 12313 / 63461 Next

Sporadic error in kalmanSmooth (PR#6738)

On Sun, 4 Apr 2004 16:24:22 +0200 (CEST), Kjetil Halvorsen wrote:

            
Found it.  The entries in the Mt array in KalmanSmooth weren't being
initialized in the cases where data was missing.  The value was being
multiplied by zero, so if it was numerically valid junk it had no
effect, but if it happened to be a NaN, the multiplication gave
another NaN.

I'll commit a patch.

Duncan Murdoch