Tuesday, December 10, 2013

gnls error, approximate covariance matrix for parameter estimates not of full rank

fm.b0.5 <- gnls( b0.5 ~ modelBlack0.5(v,w) , start = list( v = 0.0075, w=0.5)  );
Error in gnls(b0.5 ~ modelBlack0.5(v, w), start = list(v = 0.0075, w = 0.5)) :
  approximate covariance matrix for parameter estimates not of full rank

Possible problem:
There's a symmetry in the model, probably due to dilution bias. 
I manually corrected symmetry in the data, but still see the same error. 
Further scrutinizing led the discovery of a typo. 
 
Bug:
A typo in assignment that gave 'NA' bot b.05.max.

Ref:
 http://en.wikipedia.org/wiki/Rank_%28linear_algebra%29
https://svn.r-project.org/R-packages/trunk/nlme/R/gnls.R
https://stat.ethz.ch/pipermail/r-help/2003-April/032550.html

No comments:

Post a Comment