mlts 2.0.1

mlts 2.0.1

Bug Fixes

• Selection of true values for data generation in mlts_sim() was spurious for residual SDs of outcome variables. Now set to 0.5.
• Between-level standardization using mlts_standardized() failed for models with multiple time-invariant covariates.
• mlts_paths() did not display time-invariant covariates used as outcome predictors

mlts 2.0.0

New Features

• Deprecated mlts_model_paths() and replaced with mlts_paths().
• Inclusion of non-lagged (i.e., same time) directed effects via the incl_t0_effects argument in mlts_model(). Note that model formulas is not supported for these types of models at this point.
• Inclusion of interaction effects on the dynamic within-level via the incl_interaction_effects argument in mlts_model(). Note that obtaining model formulas is not supported for these types of models at this point.
• Multiple group analyses by specifying the group argument in mlts_model().
• Loosen the previous restriction of allowing only bivariate VAR models with random innovation covariances. Models are still limited to the incorporation of only one latent factor capturing the innovation covariance, however, additional constructs can be added under the assumption of independent innovations.
• Added option to include exogenous time-varying covariates (i.e., without latent-mean centering) using the is_exogenous argument in mlts_model()
• Add option to estimate models that assume censoring on the manifest indicators using the censor_left or censor_right arguments in mlts_model().
• Provide option to introduce equality constraints on loading parameters across levels in mlts_model() using equal_loads_levels = TRUE (i.e., fix between-level loading parameter of an indicator to match the respective within-level loading parameter).
• Obtain posterior predictive samples using mlts_posterior_sample()
• Run posterior predictive checks mlts_pp_check() (still considered developmental).
• Option to limit the number of consecutive missing values in the a subjects' time series to avoid excessive numbers of NAs thereby improving model estimation times.
• mlts_fit() now runs a small check to ensure that observed grand-means of ts-variables match prior distributions of the respective fixed effects. For multiple-indicator models a simple message is printed if any mean values fall outside the range of -10 to 10.

Changes

• Shortened column names in summary output
• Reordered sections in summary output
• Redundant constraint removed from re-transformed (i.e., exponentiated) individual variances in stan models
• When using tinterval in mlts_fit() now slices rows with NAs on all ts-variables prior to the creation of the time grid
• True values of population parameters using mlts_sim() with default = TRUE changed slightly.

mlts 1.0.0

• Initial CRAN submission.