Fit a generalized linear mixed model

glmm() validates the R-side family/link request, compiles the model formula, and delegates the numerical fit to the upstream Rust GeneralizedLinearMixedModel. The default method = "pirls_profiled" is the labelled fast-PIRLS path. method = "joint_laplace" uses the upstream labelled joint Laplace route (fast = FALSE, nAGQ = 1) backed by the native dependency-light optimizer in this vendored build.

Usage

glmm(
  formula,
  data,
  family,
  random = NULL,
  weights = NULL,
  offset = NULL,
  subset = NULL,
  na.action = na.omit,
  contrasts = NULL,
  method = c("pirls_profiled", "joint_laplace"),
  nAGQ = 1L,
  inference = c("auto", "none", "asymptotic", "bootstrap"),
  control = mm_control(),
  ...
)

Arguments

formula: A two-sided lme4-style formula.
data: A data.frame.
family: A supported GLMM family object or family constructor. The certified 1.0 surface is: binomial() with "logit", "probit", or "cloglog" links; poisson() with "log" or "sqrt" links; Gamma() with "log" link; and negative binomial (NB2, "log" link) via mm_negative_binomial() (theta estimated, like lme4::glmer.nb()) or MASS::negative.binomial(theta) (fixed theta).
random: Reserved for the native random-effect constructor path.
weights: Optional prior weights. For binomial models these are trial counts for proportion responses; weights must be positive and finite.
offset: Optional fixed linear-predictor offset; values must be finite.
subset, na.action, contrasts: Reserved for future parity with lmm().
method: GLMM estimation method. "pirls_profiled" is the default fast-PIRLS profiled path. "joint_laplace" requests the labelled joint Laplace route and requires nAGQ <= 1. The joint route tracks the lme4 joint-Laplace reference far more closely than the profiled path on high-baseline models, at a higher optimizer cost; cap that cost with mm_control(max_feval = ). The default profiled path is not glmer's estimator and its coefficients do not match glmer() exactly; when method is left at its default, glmm() emits an informational notice to that effect (suppress with mm_control(verbose = -1)). Use method = "joint_laplace" for glmer-equivalent estimates.
nAGQ: Number of adaptive Gauss-Hermite quadrature points. 1 is the Laplace setting. Values above 1 are allowed on the profiled path and are rejected for method = "joint_laplace" in the R wrapper.
inference: Requested inference method.
control: A list from mm_control().
...: Reserved for future use.

Value

An object of class mm_glmm, also inheriting from mm_fit and mm_compiled.

Details

Optimization runs inside a single native call with no progress output: the pre-fit explanation block (when verbose >= 0) is the last thing printed before the fitted result returns, and the call cannot be interrupted from R. Every optimizer budget is bounded, so fits always terminate; runtime on large problems is governed by mm_control(max_feval = ).

Examples

set.seed(1)
df <- data.frame(
  y = rbinom(120, 1, 0.5),
  x = rnorm(120),
  g = factor(rep(seq_len(12), each = 10))
)
fit <- glmm(y ~ x + (1 | g), df, family = binomial(),
            control = mm_control(verbose = -1))
fixef(fit)
#> (Intercept)           x 
#> -0.10067766  0.01797541 
# glmer-equivalent (joint Laplace) estimates:
fit_joint <- glmm(y ~ x + (1 | g), df, family = binomial(),
                  method = "joint_laplace",
                  control = mm_control(verbose = -1))
fixef(fit_joint)
#> (Intercept)           x 
#> -0.10067766  0.01797541