dgirt and dgmrp make calls to stan with
the Stan code and data for their respective models.
dgirt(shaped_data, ..., separate_t = FALSE, delta_tbar_prior_mean = 0.65,
delta_tbar_prior_sd = 0.25, innov_sd_delta_scale = 2.5,
innov_sd_theta_scale = 2.5, version = "2017_01_04",
hierarchical_model = TRUE, model = NULL)dgmrp(shaped_data, ..., separate_t = FALSE, delta_tbar_prior_mean = 0.65,
delta_tbar_prior_sd = 0.25, innov_sd_delta_scale = 2.5,
innov_sd_theta_scale = 2.5, version = "2017_01_04_singleissue",
model = NULL)
Output from shape.
Further arguments, passed to stan.
Whether smoothing of estimates over time should be
disabled. Default FALSE.
Prior mean for delta_tbar, the normal
weight on theta_bar in the previous period. Default 0.65.
Prior standard deviation for delta_bar.
Default 0.25.
Prior scale for sd_innov_delta, the Cauchy
innovation standard deviation of nu_geo and delta_gamma.
Default 2.5.
Prior scale for sd_innov_theta, the Cauchy
innovation standard deviation of gamma, xi, and if
constant_item is FALSE the item difficulty diff. Default
2.5.
The name of the dgo model to estimate, or the path to a
.stan file. Valid names for dgo models are "2017_01_04",
"2017_01_04_singleissue". Ignored if argument model is used.
Whether a hierarchical model should be used to
smooth the group IRT estimates. If set to FALSE, the model will return raw
group-IRT model estimates for each group. Default TRUE.
A Stan model object of class stanmodel to be used in
estimation. Specifying this argument avoids repeated model compilation. Note
that the Stan model object for a model fitted with dgirt() or
dgmrp() can be found in the the stanmodel slot of the resulting
dgirt_fit or dgmrp_fit object.
A dgo_fit-class object that extends
stanfit-class.
The user will typically pass further arguments to stan
via the ... argument, at a minimum iter and cores.
By default dgirt and dgmrp override the
stan default for its pars argument to specify
typical parameters of interest. They also set iter_r to 1L.
Important: the dgirt model assumes consistent coding of the polarity
of item responses for identification.