Movatterモバイル変換

mlts 2.0.1

Bug Fixes

• Selection of true values for data generation inmlts_sim() was spurious for residual SDs of outcomevariables. Now set to 0.5.
• Between-level standardization usingmlts_standardized()failed for models with multiple time-invariant covariates.
• mlts_paths() did not display time-invariant covariatesused as outcome predictors

mlts 2.0.0

New Features

• Deprecatedmlts_model_paths() and replaced withmlts_paths().
  
• Inclusion of non-lagged (i.e., same time) directed effects via theincl_t0_effects argument inmlts_model(). Notethat model formulas is not supported for these types of models at thispoint.
• Inclusion of interaction effects on the dynamic within-level via theincl_interaction_effects argument inmlts_model(). Note that obtaining model formulas is notsupported for these types of models at this point.
• Multiple group analyses by specifying thegroupargument inmlts_model().
• Loosen the previous restriction of allowing only bivariate VARmodels with random innovation covariances. Models are still limited tothe incorporation of only one latent factor capturing the innovationcovariance, however, additional constructs can be added under theassumption of independent innovations.
  
• Added option to include exogenous time-varying covariates (i.e.,without latent-mean centering) using theis_exogenousargument inmlts_model()
• Add option to estimate models that assume censoring on the manifestindicators using thecensor_left orcensor_right arguments inmlts_model().
  
• Provide option to introduce equality constraints on loadingparameters across levels inmlts_model() usingequal_loads_levels = TRUE (i.e., fix between-level loadingparameter of an indicator to match the respective within-level loadingparameter).
  
• Obtain posterior predictive samples usingmlts_posterior_sample()
• Run posterior predictive checksmlts_pp_check() (stillconsidered developmental).
  
• Option to limit the number of consecutive missing values in the asubjects’ time series to avoid excessive numbers ofNAsthereby improving model estimation times.
• mlts_fit() now runs a small check to ensure thatobserved grand-means ofts-variables match priordistributions of the respective fixed effects. For multiple-indicatormodels a simple message is printed if any mean values fall outside therange of -10 to 10.

Changes

• Shortened column names insummary output
• Reordered sections insummary output
• Redundant constraint removed from re-transformed (i.e.,exponentiated) individual variances in stan models
• When usingtinterval inmlts_fit() nowslices rows withNAs on allts-variables priorto the creation of the time grid
• True values of population parameters usingmlts_sim()withdefault = TRUE changed slightly.

mlts 1.0.0

• Initial CRAN submission.