Resolve selectors against the same Gaussian or binomial design builders used
by drmTMB(). Models require complete data and unit weights. Gaussian
identity models can include one independent mean random intercept; other
random effects and random-effect SD formulas are not supported with priors.
A Gamma SD prior identifies the exact group and intercept and reports shape
and rate on the natural SD scale. The native and Julia engines support this
prior for the Gaussian mean random-intercept model described above.
The Julia bridge does not yet accept factor predictors in this mixed model.
No automatic centring or scaling is performed.
Usage
drm_resolve_priors(formula, family = stats::gaussian(), data, priors)Arguments
- formula
A
drm_formula()orbf()specification.- family
Gaussian identity or binomial logit family.
- data
A data frame with complete model variables.
- priors
A collection created by
drm_priors().
Value
A data frame identifying the resolved targets. Fixed-effect rows
report the distributional parameter, coefficient, matrix column, prior
mean, SD, precision and selection reason. An SD-prior row reports the
distributional parameter, group, random-effect block, coefficient,
SD scale, shape and rate. Mixed results combine these fields, with
unavailable fields recorded as NA.
Examples
d <- data.frame(y = c(1, 2, 1, 3), x = c(0, 1, 2, 3))
p <- drm_priors(drm_prior_normal("mu", "x", sd = 2))
drm_resolve_priors(bf(mu = y ~ x, sigma = ~ 1), gaussian(), d, p)
#> dpar coefficient matrix_column mean sd precision reason kind
#> 1 mu x 2 0 2 0.25 exact column normal
d$group <- factor(c(1, 1, 2, 2))
drm_resolve_priors(bf(mu = y ~ x + (1 | group), sigma = ~ 1), gaussian(), d, p)
#> dpar coefficient matrix_column mean sd precision reason kind
#> 1 mu x 2 0 2 0.25 exact column normal