summary.fitmixture: Fixed a bug so the function
correctly returns match.prob instead of hs,
and improved standard error indexing to use regex
(grepl("gamma", ...)) rather than hardcoded indices.stop() so they throw errors, and fixed minor grammar issues
in the error messages.fit_mixture_cox: Fixed a condition length warning by
correcting the initialization check
(initbeta[1] != "default") and resolved a variable naming
inconsistency (betacur to beta_cur).predict.fitmixture: Added support for the
newdata argument. Predictions are now computed from the
reconstructed model matrix and estimated coefficients when
newdata is provided.fit_mixture_gaussian,
fit_mixture_glm): Corrected derivations to evaluate across
all observations (rather than just mismatched ones). Safe matches are
now assigned probabilities of 1 with derivatives of 0.fit_mixture_glm): Fixed
analytical first and second derivatives for binomial and gamma families
within the internal fymu_all_GLM helper function.match.prob returns the calculated
observation weights (pcur) instead of the linear predictor
probabilities (hs) across all fitting functions.DESCRIPTION, README.md,
and compiled .Rd files.m.prob: documentation to “posterior correct match
probabilities for observations given parameter estimates”fit_mixture_gaussian, fit_mixture_cox, and
fit_mixture_glm: replace sum(Delta == 1) == n, which can be
true when using a non-default (not intercept-only) mformula.fit_mixture: data can be a data.frame or list.fit_mixture: update ifelse() statements used to define
fy, initbeta, and initgamma when the control arguments are
specified.CRAN Release: 06-07-2024
DESCRIPTION: arXiv link re-formatted in the DESCRIPTION file
fit_mixture_gaussian: reformatted g function so
function input vs. y is clearer.
fit_mixture_cox: replaced match(y,times) to improve
handling of survival time ties.
CRAN Release: 03-05-2024
Initial version.