NEWS.md
estimate.mlm, IC.mlm, pars.mlm, estimate.array, estimate.data.frame
Print method for tabular data (matrix, data.frame)merge now supports regular expressionsIC returns row-names (default id) as obtained from model.matrix or similarparameter.estimate method to extract matrix with estimates, standard errors, and confidence limits from and estimate object (coefmat element)'-'.estimate and pairwise.diff
cv and bootstrap
parameter.lvm now automatically removes previously variables in the lvm object with same name as new added parameters.Print function deals more gracefully with non-rectangular objectsstack.estimate (wrong stand-errors in twostage since version 1.7.0)weibull.lvm and coxExponential.lvm now uses default parametrizations similar to rweibull, rexp. weibull.lvm now has arguments “intercept”,“sigma” that directly relates to the accelerated failure time formulation.gof, lava.tobit are removed from Suggested packages.estimate with clustered observations.uniform.lvm(value=...).cv method moved to the ‘targeted’ packageIC method that returns influence function of a model object. The iid argument iid of the estimate method is now replaced with an argument IC (with a user supplied matrix this must now be the actual influence function and not the sample-size scaled version returned by the iid method).regression("y", value=function(x) x) did not work. merge.estimate now works without IC elementmultinomial.lvm, none.lvm, constant.lvm, id.lvm.regression, regression.lvm: the ‘value’ argument can now be a (non-linear) function specifying the functional relationship between outcomes and covariates (for simulation with the sim method).intervention method for applying interventions on lvm-objectsprogressr library (enabled with progressr::handlers(global=TRUE)).future::plan("multicore").plot_region function for adding confidence regions to plots.idplot: now accepts matrix or data.frame as 1st argument. New argument: return.data.cv: rmse output fixed. score: Fixed bug for linear Gaussian model with argument ‘indiv=TRUE’. estimate.formula: call object initialized correctly. plot.lvm: ‘noplot’ argument now works with all plot engines.parameter(m,x) now returns an lvm object and not just xcomplik when used with censored variables (Surv objects).plot.sim: ‘rug’ argument is now by default FALSE and ‘auto.layout’ disabled when nr=nc=1.sequence.lvm has been renamed to Sequence.lvm. The function binary.lvm is now an alias of ones.lvm.wkm). Gaussian mixture models (mvnmix) are now initialized by default using kmeans++.sim method implemented for mvnmix models.lava::NR) used a numerical approximation of the Hessian even when submitted as attribute to the objective function.binomial.rd, binomial.rr. Base on new hook ‘simulate_multiple_inputs’ which allows the distribution to depend non-linearly on multiple different input variables.sim.lvm: ‘X’ argument can now fix (manipulate) any variable and not only exogenous variables.sim.default updated (the ‘estimate’ argument can now be a list with each element being the estimate position and optionally standard error and true value).mixture(...,names=TRUE) and set with mixture(...,control=list(start=...))).wkm.twostageCV: estimation of mixture models are now parallelized if mc.cores>1.twostageCV
mixture methodtwostageCV: cross-validate two-stage estimatorlava::rmvn0, lava::dmvn0)merge.lvm now correctly handles fixed covariance parameterssim.as method. plot.sim method now by default only plots density estimatesmixture). Fast version requires ‘mets’ packages; Gaussian mixture models (mvnmix); weighted k-means (km)estimate.default: ‘keep’, ‘use’ arguments can be specified as regular expressions (with argument regex=TRUE). Summary method now returns Wald test (null: all parameters being zero).makemissing: seed argument added.Grep, Na2x, x2NA, wait, waitclick, rotation, Rot2d, Rot3d
na.pass0: returns data.frame with original number of rows but with zeros (or first level of factors) in the rows with missing data.stack: ‘weights’ argument renamed to ‘propensity’. If propensity=TRUE, the first argument (model) will be treated as propensity score model (glm) and ‘predict’ method will be used for the predictions.estimate.formula now by default wraps glm such that the iid method return matrix of same size as full data (with zero rows where data are missing).complik) refactored + new example. ordinal method now cleans up properly when variables are removed (rmvar, subset).twostage: fixed for mixture model (class ‘lvm.mixture’). New help page + examples. Predict function updated (newdata argument where covariate levels can be specified).confpred
%++% for function compositonsummary.effects methods with mediation proportion in the outputremove.hooks (see example ordinal.lvm)sim.lvm allowing both vectorized and non-vectorized functionsnonlinear method. Estimation via the twostage function.cv (and csplit function for creating random sets).complik) updatedsim.default: new argument ‘arg’ passed on to simulation functionsim.default: new argument ‘iter’. If TRUE the iteration number is passed to function call as first argument (default FALSE)estimate.default: Wildcards/global expressions can now be used for specifying contrasts based on the syntax of the functions contr, parsedesign. See examples on the help-page. The argument transform.ci has been renamed to back.transform.revdiag: dimnames are keptCombine: output updatedforestplot: point estimates shown by defaultbackdoor now works without conditioning set (yields all possible conditioning sets)spaghetti: trend.formula can now contain a factor statement on the rhssim method now has a seed argumentNR).diagtest updated.dsep: check for d-separation (conditional independence). backdoor: check backdoor criterion of a graph (lvm-object). adjMat: return adjaceny matrix. edgeList: return edge list. ancestors: return ancenstors of nodes. descendants: return descendants of nodes.path(...,all=TRUE)
~~ instead of ,. Applies to setting starting values in estimate, parameters in sim,compare,estimate,… To use the old syntax set lava.options(symbol=c("~",",")).layout argument added to lava.options (default ‘dot’)plot.engine argument added to plot methods.bootstrap.lvmfit now default returns original estimates.print, transform methods updated (transform output).+ operator overloaded for lvm and estimate objects (merge).complik.riskcomp, rdiff, rratio, …simulate method. If FALSE the latent variables are dropped from the returned data.frame.modelsearch by default now shows both directional or undirectional associations (type=‘all’ vs type=‘cor’).sim.default now stores timings. New print functions (data.table like output).sim function, for instance setting parameter values for the simulation only once: m <- sim(m,p=p,...), with faster subsequent calls sim(m,n=n).estimate.default can now simulate p-values (‘R’ argument). Returns an object which can also be used as input for estimate.NR optimization with back-tracing; fixed matrices.lvm when called without variance parameters; fixed a bug in r-square computations.contr.