设置参数估计的界限 - R package 盛况

Question

We are fitting a model in using the pomp package in R and want to place some bounds on the parameters we are estimating because they are exploring a space that is non-sensical and we want to keep it out of that parameter space. Our current "best option" is to transform the parameter in the model code in C in a way that limits it between the upper and lower bound. So for example, if we want to limit parameter beta between 0 and 0.05, we could logit transform the estimated parameter so it is between 0 and 1 and multiply this by the upper bound. This is what the code would be in the sections in our code where we define the model in C snippets:

beta = ((exp(est_beta)/(1+exp(est_beta))*max_beta;

where est_beta is the parameter estimation in the space from -Inf to Inf and max_beta is the upper bound.

So my question is, is there a more elegant way to either 1) impose bounds on parameter estimation in a pomp model or 2) impose bounds or create a truncated variable in C ?

Answer 1

pomp does have an explicit facility for defining these transforms outside of the core model code: https://kingaa.r-universe.dev/manual/pomp.html#parameter_trans

Note that estimation software frequently uses transforms such as logit at a low level to implement bounds, so I would not call this inelegant. (eg,: Stan: https://mc-stan.org/docs/2_22/reference-manual/variable-transforms-chapter.html )

However, it may make more sense to use a strong prior to avoid sampling nonsensical space.