LMMstar.options: Global options for LMMstar package

In bozenne/repeated: Repeated Measurement Models for Discrete Times

Global options for LMMstar package

Description

Update or select global options for the LMMstar package.

Usage

LMMstar.options(..., reinitialise = FALSE)

Arguments

...	options to be selected or updated
reinitialise	should all the global parameters be set to their default value

Details

The options are:

• adj.method [character vector]: possible methods to adjust for multiple comparisons. NOT MEANT TO BE CHANGED BY THE USER.

• backtransform.confint [logical]: should variance/covariance/correlation estimates be back-transformed when they are transformed on the log or atanh scale. Used by confint.

• columns.anova [character vector]: columns to ouput when using anova with argument ci=TRUE.

• columns.confint [character vector]: columns to ouput when using confint.

• columns.summarize [character vector]: columns to ouput when displaying descriptive statistics using summarize.

• columns.summary [character vector]: columns to ouput when displaying the model coefficients using summary.

• df [logical]: should approximate degrees-of-freedom be computed for Wald and F-tests. Used by lmm, anova, predict, and confint.

• drop.X [logical]: should columns causing non-identifiability of the model coefficients be dropped from the design matrix. Used by lmm.

• effects [character]: parameters relative to which estimates, score, information should be output.

• min.df [integer]: minimum possible degree-of-freedom. Used by confint.

• method.fit [character]: objective function when fitting the Linear Mixed Model (REML or ML). Used by lmm.

• method.numDeriv [character]: type used to approximate the third derivative of the log-likelihood (when computing the degrees-of-freedom). Can be "simple" or "Richardson". See numDeriv::jacobian for more details. Used by lmm.

• n.sampleCopula [integer]: number of samples used to compute confidence intervals and p-values adjusted for multiple comparisons via "single-step2". Used by confint.Wald_lmm.

• optimizer [character]: method used to estimate the model parameters. Either "FS", an home-made fisher scoring algorithm, or a method from optimx:optimx like "BFGS"or Nelder-Mead.

• param.optimizer [numeric vector]: default option for the FS optimization routine: maximum number of gradient descent iterations (n.iter), maximum acceptable score value (tol.score), maximum acceptable change in parameter value (tol.param), method to initialize the correlation parameters (init.cor).

• pool.method [character vector]: possible methods to pool estimates. NOT MEANT TO BE CHANGED BY THE USER.

• precompute.moments [logical]: Should the cross terms between the residuals and design matrix be pre-computed. Useful when the number of subject is substantially larger than the number of mean paramters.

• sep [character vector]: character used to combined two strings of characters in various functions (lp: .vcov.model.matrix, k.cov/k.strata: .skeletonK, pattern: .findUpatterns, rho.name/rho.strata: .skeletonRho, reformat: .reformat ).

• trace [logical]: Should the progress of the execution of the lmm function be displayed?

• tranform.sigma, tranform.k, tranform.rho: transformation used to compute the confidence intervals/p-values for the variance and correlation parameters. See the detail section of the coef function for more information. Used by lmm, anova and confint.

• type.information [character]: Should the expected or observed information ("expected" or "observed") be used to perform statistical inference? Used by lmm, anova and confint.

Value

A list containing the default options.

bozenne/repeated documentation built on July 16, 2025, 11:16 p.m.