Nothing
hbam()
with the argument extra_pars = "Y_pred"
: HBAM, HBAM_MULTI, HBAM_NF, HBAM_MULTI_NF, BAM.prep_data()
function now prints a summary of the input and output data to the console. It also throws an error if the selection criteria for inclusion in the analysis are too strict to retain any respondents.plot_stimuli()
has been slightly improved.theta
and rho
have been made narrower to make the models more robust to situations with extremely scarce data. rho
(which appears in all heteroskedastic models) has been made narrower to reduce the risk of divergent transitions when there are few observations per stimuli (which could be the case for e.g. expert surveys). mu_beta
and mu_alpha
in MULTI-type models and the priors on sigma_beta
and sigma_alpha
in HBAM-type models have also been made narrower to reduce the risk of sampling issues.get_est()
function now takes the logical argument format_orig
, which if TRUE
makes the function return posterior summaries for individual-level parameters in a format that matches the rows in the original dataset.hbam()
and fbam()
functions now store the input data within the returned objects, which allows simplifying the interfaces for other functions like get_est()
and plot_over_self()
. As a result, the plot_over_self()
function no longer requires a data argument.prep_data()
, hbam()
, and fbam()
functions now allow users to not supply self-placements. In this case, no meaningful respondent positions will be estimated, but all other parameters are unaffected.prep_data()
function now allows the group_id
argument to take various forms, such as factor or character. It also allows missing values in the group_id
vector and will drop respondents who do not have a valid group_id
. Missing values would previously generate an uninformative error message.prep_data()
will throw a warning.prep_data()
function now identifies the left and right poles for the BAM model as the stimuli with the most non-NA observations on each side of the center. This can be advantageous when analyzing datasets where some stimuli have a much higher number of valid observations than others. hbam_cv()
function no longer uses parallel::mclapply()
for parallel computation as the latter relies on forking, which is not available on Windows. hbam_cv()
has been revised to work with the future
package, where the user decides the computational strategy and options are available for parallel computation on all systems. hbam_cv()
has been changed to comply with the standards of the loo
package. The function now returns a list with classes kfold
and loo
. This allows the user to compare estimated ELPDs and obtain standard errors for their differences via loo::loo_compare()
. show_code()
shows Stan code for any model in the package.group_id
. The model gives each group separate hyperparameters for the locations of the prior distributions for the shift and stretch parameters. Rather than shrinking the estimates toward the mode for the whole dataset, this model shrinks the estimates toward the mode for the group. The vectors of hyperparameters are called mu_alpha
and mu_beta
and are constructed to have means of 0. The scales of the priors on these hyperparameters can be set by the user via the arguments sigma_mu_alpha
and sigma_mu_beta
. One potential use for this model is to supply self-placements as group_id
, and thus give each self-placement group its own prior distribution for the shift and stretch parameters.sigma_alpha
and sigma_beta
, and set the scales of the priors on mu_alpha
and mu_beta
via the arguments sigma_mu_alpha
and sigma_mu_beta
.Any scripts or data that you put into this service are public.
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.