truncated normal distribution specification in debinfer_par

[motyocska]

Hello,

would you have an example on how to specify a truncated normal distribution for a prior in debinfer_par? looked through repository but could not find example... For example, how would I specify debinfer_par for a parameter with normal distribution, mean of 5, sd of 2, lower end of 1 and upper end of 10? Tried the following but error message pops for unused variables:

debinfer_par(name = "par_a", var.type = "de", fixed = FALSE,value = 5, prior = "norm", hypers = list(mean = 5, sd = 2,a = 1, b = 10),prop.var = c(4), samp.type = "rw")

and to confirm, the prop.var variable for my example would be squared 2 = 4, the variance?
much appreciate the help,

Andras

[pboesu]

Hi Andras,

truncated distributions can be implemented using the truncdist package and the relevant dtrunc etc. functions: https://search.r-project.org/CRAN/refmans/truncdist/html/dtrunc.html i.e. the actual distribution name becomes a hyper parameter, passed onto the spec argument of dtrunc

so your parameter specification should look something like

debinfer_par(name = "par_a", var.type = "de", fixed = FALSE,value = 5, prior = "trunc", hypers = list(spec = "norm", mean = 5, sd = 2,a = 1, b = 10),prop.var = c(4), samp.type = "rw")

[pboesu]

And worked examples with truncated priors and likelihoods are given in this paper: https://doi.org/10.1016/j.seares.2018.07.014

Open Access preprint here
Code repository here