[R] When using Rdonlp2 to do a MLE, I got some "system is exactly singular" error

Hsiao-nan Cheung niheaven at gmail.com
Tue Mar 8 15:01:02 CET 2011


Hi,

Recently I use Rdonlp2 to do a MLE of Markov-switching with TVTP, but
during the optimization, there is always a "Error in
solve.default(...) : Lapack routine dgesv: system is exactly singular"
error, what's wrong? When searching on the web, someone said it is
because the Hessian matrix, and could be fixed by modified Marquardt
algorithm, and is this algorithm realized in R? Or what else could I
do?

Thanks!

Hsiao-nan Cheung



More information about the R-help mailing list