The hardware and bandwidth for this mirror is donated by dogado GmbH, the Webhosting and Full Service-Cloud Provider. Check out our Wordpress Tutorial.
If you wish to report a bug, or if you are interested in having us mirror your free-software or open-source project, please feel free to contact us at mirror[@]dogado.de.
Corrected binary home-field effects at neutral sites and retained the score intercept when overtime is modeled without home-field effects.
Corrected Poisson likelihood constants for zero and one, stabilized probit derivatives in the negative tail, and safeguarded conditional-mode and fixed-effect Newton updates.
Corrected fully exponential joint-model covariance and fixed-effect updates to include both response families. Final ratings are now recomputed at the returned model parameters.
Unified exact Gaussian ML/REML computations. REML without a margin home effect now omits the fixed effect and its uncertainty. Corrected likelihood constants, covariance scores, and fixed-model residual degrees of freedom.
Enabled Hessian output for N.mov and corrected Hessians to differentiate the negative likelihood. Gaussian REML Hessians cover covariance parameters; fixed-effect uncertainty remains in N.output.
Corrected N.mov predictions and added explicit errors for invalid input, unknown teams, unidentifiable fixed effects, and unsupported overtime options. Preserved the historical paired-binary treatment of ties and documented its interpretation.
Added final logLik, iterations, converged, and approximation diagnostics for every method. P1 now reports its game-effect variance in G, and Poisson parameters have names without requesting a Hessian. verbose=FALSE suppresses progress output.
Moved all function and dataset documentation into roxygen source comments, regenerated with roxygen2 8.1.0, and added testthat coverage of all methods and numerical/statistical identities.
These corrections can change previous estimates, including default home-field and fully exponential fits. Public function arguments and defaults are unchanged.
Fixed handling of method = "N.mov" with
home.field = FALSE. In previous versions this option was
silently ignored because N_mov() forced
home.field = TRUE internally.
Fixed related no-home-field parameter indexing in the
N_mov() and NB_mov() score/Hessian
helpers.
Fixed no-home-field binary fixed-effect assignment in the
PB_cre() Hessian helper.
Added neutral-site/home-field smoke coverage. Short regression
coverage is run during package checks; the full option matrix can be run
locally with NOT_CRAN=true.
Refreshed user documentation, examples, references, and citation metadata.
Default home.field = TRUE behavior is intended to be
unchanged. Prior results are expected to be affected only for calls
using method = "N.mov", home.field = FALSE.
Added paper reference about potential bias in the home field effect estimates.
Changed default for REML.N to
TRUE.
method = "N".method = "N.mov".REML.N argument.Changed option method = "Richardson" to
method = "simple" for the jacobian() function
from package numDeriv.
Changed the sign of the Hessian.
Now reports scaled conditional residuals for models N and NB.
These binaries (installable software) and packages are in development.
They may not be fully stable and should be used with caution. We make no claims about them.
Health stats visible at Monitor.