probit: Probit analysis with random effects
In bomarkussen/probit: Item response analysis with random effects.

Description Usage Arguments Value Note

View source: R/probit.R

Separate analysis over items.

probit(
  fixed,
  random,
  subject.name = "id",
  dependence = "marginal",
  Gamma = NULL,
  item.name = "item",
  response.name = NULL,
  data,
  data.long = NULL,
  mu = NULL,
  psi = NULL,
  B = 300,
  BB = 50,
  maxit = 20,
  sig.level = 0.6,
  verbose = 0
)

`fixed`	Model formula for the fixed effect, where multivariate responses may be given additively on the left hand side. Responses must be ordered factors.
`random`	List of formulas for the random effects. Models are fitted on first appearance observations.
`subject.name`	Character string with name of categorical variable encoding subjects.
`dependence`	Text string (`"marginal" or "joint"`) deciding whether random effects are assumed independent or with a common joint normal distribution. Default: `dependence="marginal"`.
`Gamma`	Choleskey factor for initial variance of random effects. If `Gamma=NULL` then initialized at identity matrix. Default: `Gamma=NULL`.
`item.name`	Character string with name of generated variable identifying the items. Defaults to `item.name="item"`.
`response.name`	Character string with name of generated variable containing the responses. Defaults to `response.name=NULL`, which corresponds to adding `".value"` to `item.name`.
`data`	Date frame with data on the wide format.
`data.long`	Possible additional long format data frame with item level explanatory variables. Presently not implemented!
`mu`	Matrix (no subjects, q) of initial estimates for mu.
`psi`	Matrix (no subjects, q*(q+1)/2) of initial estimates for psi.
`B`	Number of simulations in minimization step. Default: `B=300`.
`BB`	Number of simulations per subject in maximization step. Default: `BB=50`.
`maxit`	Maximal number of minimization-maximization steps. Default: `maxit=20`.
`sig.level`	Significance level at which the iterative stochastic optimizations will be stopped. Defaults to `sig.level=0.60`.
`verbose`	Numeric controlling amount of convergence diagnostics. Default: `verbose=0`. Level `1` gives progress bar and convergence statistics. Level `2` also gives diagnostics for maximization step. Level `3` also gives graphical diagnostics for minimization step.

probit-class object.

A data frame must be provided, i.e. the data option is not optional. Variables that coincide with random effects, item identifier or internal name for item response will be removed from data, and an warning will be issued.

The minimization step is implemented via future_map. This implies that the user may activate parallel computations by calling plan, e.g. future::plan("multisession", workers = 4).

bomarkussen/probit documentation built on April 3, 2021, 7:38 p.m.