Optimizing an external function of the predictors


#1

It may be that I have not entirely understood what I have researched online but is there an example of how to include a user defined function in brms, where the function is of the predictor variable(s) and there is a parameter to be optimized. The function f() consumes one of the predictor variables in a regression and there is a parameter p1 that I need to optimize during fitting. Is this possible?

y~f(x,p1)


#2

Specifically, if it helps, I am looking to optimize the adstock function:

adstock_loop<-function(x,rate)
{
adstock_x<- vector(mode=“numeric”, length=length(x))
adstock_x[1]<-x[1]

for (i in 2:length(x))
{

adstock_x[i]<-x[i]+rate*adstock_x[i-1]

}

return(adstock_x)
}


#3

Is the best way to generate the stan code and then try and alter it to add this function in?


#4

Could be one option if you don’t want to write the Stan code from scratch.