Likelihood profiles

Wraps TMB::tmbprofile() per parameter. The returned objects have plot() and confint() methods (from TMB).

Usage

# S3 method for class 'frmtmb_fit'
profile(fitted, parm, ...)

Arguments

fitted

A frmtmb_fit.

parm

Parameter names or indices. Required; profiling is not free, so there is no all-parameters default. The names are the ones confint() takes, in any of its three spellings: internal (theta_1, tarsus_(Intercept)), parenthesis-free (tarsus_Intercept), or a one-to-one natural-scale alias (sd_dam__Intercept). The profile is of the internal parameter either way - a log standard deviation, not a standard deviation - and the returned element keeps the internal name. For a profile of a natural-scale quantity itself, including one that mixes several parameters, use hypothesis(method = "profile").

...

Passed to TMB::tmbprofile().

Value

A tmbprofile data frame, or a named list of them when parm has length above one.

Examples

set.seed(1)
dd <- data.frame(x = rnorm(100), g = factor(rep(1:10, 10)))
dd$y <- rnorm(100, 1 + 0.5 * dd$x + rnorm(10, 0, 0.8)[dd$g], 1)
fit <- frm(bf(y ~ x + (1 | g)) + gaussian(), data = dd)

# parameter names are the confint() row names, with or without the
# parentheses, or a one-to-one natural-scale alias
rownames(confint(fit))
#> [1] "(Intercept)"       "x"                 "sigma_(Intercept)"
#> [4] "theta_1"          
pr <- profile(fit, "theta_1")
identical(profile(fit, "(Intercept)"), profile(fit, "Intercept"))
#> [1] TRUE
plot(pr)

# TMB's confint() reads the interval off the profile
confint(pr)
#>            lower     upper
#> theta -0.6835298 0.3479822

# several parameters at once return a named list
prs <- profile(fit, c("x", "theta_1"))
names(prs)
#> [1] "x"       "theta_1"