- Fixed a couple of things

- Now the eigenvalue based solutions are nicely consistent with the deSolve solutions, if enough
  output times are specified (100, sometimes more are needed, see test.R)
- Workaround for invilr not to produce NaN values so often
- Still a lot to do (see TODO)



git-svn-id: svn+ssh://svn.r-forge.r-project.org/svnroot/kinfit/pkg/mkin@30 edb9625f-4e0d-4859-8d74-9fd3b1da38cb

1 files changed, 7 insertions, 0 deletions

diff --git a/R/ilr.R b/R/ilr.R
index b53f868..389653e 100644
--- a/R/ilr.R
+++ b/R/ilr.R
@@ -40,5 +40,12 @@ invilr<-function(x) {
   for (i in 1:D) {
     z[i] <- exp(y[i])/sum(exp(y))
   }
+
+  # Work around a numerical problem with NaN values returned by the above
+  # Only works if there is only one NaN value: replace it with 1
+  # if the sum of the other components is < 1e-10
+  if (sum(is.na(z)) == 1 && sum(z[!is.na(z)]) < 1e-10) 
+    z = ifelse(is.na(z), 1, z)
+
   return(z)
 }