Specify Prior for Dispersion or Standard Deviation

Specify the mean of prior for the dispersion parameter (in Poisson and binomial models) or the standard deviation parameter (in normal models.)

Usage

set_disp(mod, mean)

Arguments

mod

An object of class "bage_mod", created with mod_pois(), mod_binom(), or mod_norm().

mean

Mean value for the exponential prior. In Poisson and binomial models, can be set to 0.

Value

A bage_mod object

Details

The dispersion or mean parameter has an exponential distribution with mean \(\mu\),

$$p(\xi) = \frac{1}{\mu}\exp\left(\frac{-\xi}{\mu}\right).$$

In Poisson and binomial models, mean can be set to 0, implying that the dispersion term is also 0. In normal models, mean must be non-negative.

If set_disp() is applied to a fitted model, it 'unfits' the model, deleting existing estimates.

See also

• mod_pois(), mod_binom(), mod_norm() Specify a model for rates, probabilities, or means

• set_prior() Specify prior for a term

• set_n_draw() Specify the number of draws

• is_fitted() Test whether a model is fitted

Examples

mod <- mod_pois(injuries ~ age:sex + ethnicity + year,
                data = nzl_injuries,
                exposure = popn)
mod
#> 
#>     ------ Unfitted Poisson model ------
#> 
#> 
#>    injuries ~ age:sex + ethnicity + year
#> 
#>   exposure = popn
#> 
#> 
#>         term  prior along n_par n_par_free
#>  (Intercept) NFix()     -     1          1
#>    ethnicity NFix()     -     2          2
#>         year   RW()  year    19         18
#>      age:sex   RW()   age    24         22
#> 
#> 
#>  n_draw pr_mean_disp var_time var_age var_sexgender
#>    1000            1     year     age           sex
#> 
mod |> set_disp(mean = 0.1)
#> 
#>     ------ Unfitted Poisson model ------
#> 
#> 
#>    injuries ~ age:sex + ethnicity + year
#> 
#>   exposure = popn
#> 
#> 
#>         term  prior along n_par n_par_free
#>  (Intercept) NFix()     -     1          1
#>    ethnicity NFix()     -     2          2
#>         year   RW()  year    19         18
#>      age:sex   RW()   age    24         22
#> 
#> 
#>  n_draw pr_mean_disp var_time var_age var_sexgender
#>    1000          0.1     year     age           sex
#> 
mod |> set_disp(mean = 0)
#> 
#>     ------ Unfitted Poisson model ------
#> 
#> 
#>    injuries ~ age:sex + ethnicity + year
#> 
#>   exposure = popn
#> 
#> 
#>         term  prior along n_par n_par_free
#>  (Intercept) NFix()     -     1          1
#>    ethnicity NFix()     -     2          2
#>         year   RW()  year    19         18
#>      age:sex   RW()   age    24         22
#> 
#> 
#>  n_draw pr_mean_disp var_time var_age var_sexgender
#>    1000            0     year     age           sex
#>