rstanarm v2.32.1

What's Changed

• Avoid error for 1-D unit_vector by @jgabry in #606
• Update .set_nreps by @avehtari in #612
• Use LTO during compilation

Full Changelog: v2.26.1...v2.32.1

rstanarm v2.26.1

This release updates rstanarm to use the latest syntax supported by rstan.

What's Changed

• wells dataset wrong variable name by @storopoli in #552
• add logo to be displayed in r-universe by @avehtari in #569
• Fix CRAN NOTEs and failures by @andrjohns in #587
• Implement posterior::as_draws() for rstanarm objects by @jgabry in #596
• Array syntax by @bgoodri in #597

Full Changelog: v2.21.3...v2.26.1

rstanarm v2.21.3

Bug fixes

• Fix bug where loo() with k_threshold argument specified would error if the model formula was a string instead of a formula object. (#454)

• Fix bug where loo() with k_threshold argument specified would error for
  models fit with stan_polr(). (#450)

• Fix bug where stan_aov() would use the wrong singular.ok logic. (#448)

• Fix bug where contrasts info was dropped when subsetting the model matrix in
  stan_glm(). (#459)

• Fix bug where stan_glmer() would error if prior_aux=NULL. (#482)

• posterior_predict() and posterior_epred() don't error with newdata for
  intercept only models by allowing data frames with 0 columns and multiple rows. (#492)

New features

• New vignette on AB testing. (#409)

• stan_jm() gains an offset term for the longitudinal submodel. (#415, @pamelanluna)

• Effective number of parameters are computed for K-fold CV not just LOO CV. (#462)

• stan_clogit() now allows outcome variable to be a factor. (#520)

rstanarm v2.21.1

rstanarm 2.21.1

• Compatible with rstan v2.21.1
• Consistent with new book Regression and Other Stories

Backwards incompatible changes

• stan_jm() is not available for 32bit Windows

• Some improvements to prior distributions, as described in detail in the vignette
  Prior Distributions for rstanarm Models and book Regression and Other Stories.
  These changes shouldn't cause any existing code to error, but default priors have
  changed in some cases:

  • default prior on intercept is still Gaussian but the way the location and
    scale are determined has been updated (#432)
  • autoscale argument to functions like normal(), student_t(), etc.,
    now defaults to FALSE except when used by default priors (default
    priors still do autoscalinng). This makes it simpler to specify non-default
    priors. (#432)

Bug fixes

• Fixed error in kfold() for stan_gamm4() models that used random argument (#435)
• Fixed error in posterior_predict() and posterior_linpred() when using newdata with
  family = mgcv::betar (#406, #407)
• singular.ok now rules out singular design matrices in stan_lm() (#402)
• Fix a potential error when data is a data.table object (#434, @danschrage)

New functions

• New method posterior_epred() returns the posterior distribution of the
  conditional expectation, which is equivalent to (and may eventually entirely
  replace) setting argument transform=TRUE with posterior_linpred(). (#432)

• Added convenience functions logit() and invlogit() that are just wrappers
  for qlogis() and plogis(). These were previously provided by the arm
  package. (#432)

v2.19.2

rstanarm version 2.19.2 is now on CRAN. Binaries have been built so it should be possible to install via

install.packages("rstanarm")

Release notes are available on the rstanarm website at http://mc-stan.org/rstanarm/news/

v2.14.1

Bug fixes

• VarCorr could return duplicates in cases where a stan_{g}lmer model used grouping
  factors with spaces. This is now fixed.
• The pairs function now also allows with group-specific parameters
• The stan_gamm4 function works better now (see issues #136, #132)
• Fix a problem with factor levels after estimating a model via stan_lm

New features

• New model-fitting function stan_betareg (and stan_betareg.fit) that uses the same likelihoods as those supported by the betareg function in the betareg package
• New choices for priors on regression coefficients: laplace, lasso, product_normal
• The hs and hs_plus priors now have new global_df and global_scale arguments
• stan_{g}lmer models that only have group-specific intercept shifts are considerably
  faster now
• Models with Student t priors and low degrees of freedom (that are not 1, 2, or 4) may work better now due to Cornish-Fisher transformations
• Many functions for priors (e.g., normal, student_t, etc.) have gained an autoscale
  argument that defaults to TRUE and indicates that rstanarm should make internal changes to the prior based on the scales of the variables. This doesn't change previous behavior, just replaces the former mechanism for controlling this (used to be the scaled argument to the now deprecated prior_options function).
• The new compare_models function is a wrapper for loo::compare that does more extensive checking that the rstanarm models being compared are compatible
• New prior_aux argument allow specifying priors for auxiliary parameters. Previously it was only possible to set the scale for the prior on auxiliary parameters (e.g., residual sd for Gaussian, shape for Gamma, etc.). With the prior_aux argument it is now possible to use exponential, normal, student_t, or cauchy for auxiliary parameters.

Deprecated arguments

• The prior_ops argument to various model fitting functions is deprecated and the prior_options function used to specify the prior_ops argument is also depredated. The functionality of prior_options has been replaced by the autoscale argument to the functions for setting priors as well as the prior_aux argument for for the prior on the auxiliary parameter of various GLM-like models.

v2.13.1

prepare for release of 2.13.1