The coefficients of each linear predictor, as a named list keyed by
distributional parameter (and by response on a multivariate fit),
each entry named by DESIGN COLUMN: fixef_by_dpar(fit)$sigma,
fixef_by_dpar(fit)$mu[["(Intercept)"]].
Details
This is the shape fixef() returned before frmtmb took brms's
summary matrix. It is kept because nothing else has it: fixef()
names its rows as brms does and flattens every predictor into one
table, and fixef(flatten = TRUE) is one vector in the confint()
spelling. A model with several linear predictors (a mixture, a
multivariate response, a nonlinear body) is the case this reads
cleanly.
It carries ESTIMATES only. For a standard error or an interval use
fixef(), whose rows are the same coefficients under brms's names.
Examples
set.seed(1)
dd <- data.frame(x = rnorm(100))
dd$y <- rnorm(100, 1 + 0.5 * dd$x, exp(0.2 + 0.1 * dd$x))
fit <- frm(bf(y ~ x, sigma ~ x) + gaussian(), data = dd)
# one entry per distributional parameter, named by design column
fixef_by_dpar(fit)
#> $mu
#> (Intercept) x
#> 0.9540022 0.5184330
#>
#> $sigma
#> (Intercept) x
#> 0.1515421 0.1026323
#>
exp(fixef_by_dpar(fit)$sigma[["(Intercept)"]])
#> [1] 1.163627
# the same coefficients under brms's names, with their errors
fixef(fit)
#> Estimate Est.Error Q2.5 Q97.5
#> Intercept 0.9540022 0.11693544 0.72481301 1.1831915
#> sigma_Intercept 0.1515421 0.07154043 0.01132541 0.2917587
#> x 0.5184330 0.13753442 0.24887050 0.7879955
#> sigma_x 0.1026323 0.09977149 -0.09291626 0.2981808