CAT_DT {cat.dt}R Documentation

CAT decision tree

Description

Generates a cat.dt object containing the CAT decision tree. This object has all the necessary information to build the tree.

Usage

CAT_DT(
  bank,
  model = "GRM",
  crit = "MEPV",
  C = 0.3,
  stop = c(6, 0),
  limit = 200,
  inters = 0.98,
  p = 0.9,
  dens,
  ...
)

Arguments

bank

data.frame or matrix of the item bank. Rows represent items, and columns represent parameters. If the model is "GRM", the first column represents the alpha parameters and the next columns represent the beta parameters. If the model is "NRM", odd columns represent the alpha parameters and even columns represent beta parameters

model

polytomous IRT model. Options: "GRM" for Graded Response Model and "NRM" for Nominal Response Model

crit

item selection criterion. Options: "MEPV" for Minimum Expected Posterior Variance and "MFI" for Maximum Fisher Information

C

vector of maximum item exposures. If it is an integer, this value is replicated for every item

stop

vector of two components that represent the decision tree stopping criterion. The first component represents the maximum level of the decision tree, and the second represents the minimum standard error of the ability level (if it is 0, this second criterion is not applied)

limit

maximum number of level nodes

inters

minimum common area between density functions in the nodes of the evaluated pair in order to join them

p

a-priori probability that controls the tolerance to join similar nodes

dens

density function (e.g. dnorm, dunif, etc.)

...

parameters of the density function

Value

An object of class cat.dt

Author(s)

Javier Rodr?guez-Cuadrado

Examples

data("itemBank")
# Build the cat.dt
nodes = CAT_DT(bank = itemBank, model = "GRM", crit = "MEPV",
               C = 0.3, stop = c(3,0.05), limit = 100, inters = 0.9,
               p = 0.9, dens = dnorm, 0, 1)

# Estimate the ability level of a subject with responses res
CAT_ability_est(nodes, res = itemRes[1, ])
# or
nodes$predict(res = itemRes[1, ])
# or
predict(nodes, itemRes[1, ])


[Package cat.dt version 0.3.1 Index]