[R-sig-dyn-mod] changing state variable in simulations

[Thomas Petzoldt]

Hi,

thanks to Daniel for the contribution. Regarding performance, "vode" (or
"bdf") is often faster than "lsoda" for stiff systems where "adams" is
faster for non-stiff systems.

For your case I get the following from a loop with 100 runs:

lsoda: 9.17s
vode:  9.33s
bdf:   9.00s
adams: 7.57s

It indicates that the system is not stiff -- not very surprising for a
system with only one state variable.


For stiff systems "bdf" and "vode" are superior in most cases and
"adams" would be the slowest. "vode" (and "radau") are candidates for
difficult cases.

Our observation is, that "lsoda" is often the second best, because it
can switch automatically between "bdf" and "adams".

Thomas