Build a biclustering reference from published field profiles

Assembles the minimal object that equate() and predict() need from a field reference profile matrix and the field of each item, so that a new form can be equated, or new respondents scored, from published values alone, without the data the model was fitted to.

Usage

as_biclustering(
  FRP,
  fields,
  category_shift = 0,
  ordered = FALSE,
  msg = c("Class", "Rank")
)

Arguments

FRP

The field reference profiles: a fields x classes matrix of correct-response probabilities for binary data, or a fields x classes x categories array of category probabilities for polytomous data (each row summing to one over the categories).

fields

The field of each item, as an integer vector named by the item labels (a factor or a character vector of field numbers is also accepted).

category_shift

Polytomous data only: the amount the response codes must be moved up to start at 1 (0 when the codes are already 1, 2, ...; 1 when they are 0-based). The CategoryShift of a fit.

ordered

Whether the profiles are order-restricted across the classes (an isotonic Ranklustering). Only the new-field diagnostic of equate() reads it.

msg

"Class" or "Rank": how the latent groups are labelled.

Value

An object of class BiclusteringReference, accepted by equate() and predict().

Examples

fit <- Biclustering(J15S500, ncls = 3, nfld = 2, method = "B", verbose = FALSE)
ref <- as_biclustering(fit$FRP, fit$FieldEstimated)
p <- predict(ref, J15S500$U[1:5, ])
p$Estimate
#> Student1 Student2 Student3 Student4 Student5 
#>        1        2        2        2        1