I’ve got a question regarding how to implement variable boundary conditions for the parameter estimates.
Namely, something like
fixef( tvpar1 = c(0, 1, ) )
fixef( tvpar2 = c(0, 0.1, tvpar1) )
OR
par3 = 3*par1 – 1
fixef( tvpar1 = c(0, 1, ) )
fixef( tvpar2 = c(0, 0.1, par3) )
Many thanks!
Chiara
Dear Chiara,
can you describe what you are trying to do ?
consider below you constrain for example Ka to be always greater than Ke by writing an appropriate stparm :
stparm(Ka = tvKe + tvKadelta )
stparm(Ke = tvKe )
fixef(tvKadelta = c(0, 1, ))
fixef(tvKe = c(, 1, ))
for a more technical overview http://mc-stan.org/users/documentation/case-studies/identifying_mixture_models.html
Exchangeable Inferences and Ordering Constraints
chiaraf
February 12, 2018, 9:51am
3
Dear smouksass1,
Thanks for your reply.
What I’d like to implement is very similar to what you described. Most of my model parameters are connected by algebraic equations, so many of them have boundaries which depend on non linear combinations of other parameter values.
See below a simplified example:
I want to estimate Base1_0, Base2_0, Base3_0, kcirc1, kcirc2 and kprol, such that
kprol < (Base1_0kcirc1 + Base2_0 kcirc2)/Base3_0
or, similarly
kprol = (Base1_0kcirc1 + Base2_0 kcirc2)/Base3_0 - Deltakprol
where Deltakprol has to be smaller than (Base1_0kcirc1 + Base2_0 kcirc2)/Base3_0
Many thanks in advance,
Chiara