Minor changes
- Allow posterior draws of
e to be used in predictions (with the flag use_e = TRUE, the new default in posterior_predict) or generate e based on posterior draws of sigma_e (obtained by setting use_e = FALSE)
Major changes
- added a second option for obtaining predictions on new data using a previously fitted model: the options are now
predict_paramedic and posterior_predict. The former uses posterior draws and a Stan algorithm to generate predictions; the latter uses posterior draws and internal R functions to generate predictions. Both are expected to perform similarly.
Minor changes
- Updated return type of
run_paramedic to class paramedic; added associated print, summary functions
Major changes
- added ability to obtain predictions on new data using a previously fitted model using
predict_paramedic
Major changes
- added ability to run analysis adjusted for batch effects by setting argument
k to be nonzero in run_paramedic (the default is zero for data collected in a single experiment)
- added new stan code for batch analysis
Minor changes
- reorganized stan code into directories, since much of the code was shared between models; the new organization uses
#include statements, and the shared code is stored in data/ (sets up data objects), parameters/ (sets up model parameters), tparameters/ (sets up transformed parameters), and model/ (sets up hierarchical model for shared parameters)
Major changes
- removed function
run_paramedic_centered; the desired behavior can now be obtained by using run_paramedic with the argument "centered = TRUE"
- removed function
no_efficiency; the desired behavior can now be obtained by using run_paramedic with arguments "alpha_sigma = 0" and "kappa_sigma = 0"
- removed several stan models by collapsing functionality into
variable_efficiency.stan and variable_efficiency_centered.stan
Minor changes
- updated all tests to reflect major changes
- added vignette discussing hierarchical model specification
- added a
pkgdown site
Minor changes
- new internal functions to reduce replicating code within key functions
- new
stan model for specifying a negative binomial distribution on V
Minor changes
- update internal
Stan models to use log densities and unconstrained parameters, then back-transform after sampling
Major changes
- allow an option to adjust for covariates
Minor changes
- allow
run_paramedic with q = q_obs
This is the initial package release!