fit.irtree_model: Fit an ItemResponseTrees model

In ItemResponseTrees: IR-Tree Modeling in 'mirt', 'Mplus', or 'TAM'

Description

This function takes a data frame and an object of class irtree_model and runs the model in either mirt, Mplus, or TAM.

Usage

## S3 method for class 'irtree_model'
fit(
  object = NULL,
  data = NULL,
  engine = c("mirt", "mplus", "tam"),
  ...,
  link = c("logit", "probit"),
  verbose = interactive(),
  control = NULL,
  improper_okay = FALSE
)

Arguments

object	Object of class irtree_model. See irtree_model for more information.
data	Data frame containing containing one row per respondent and one column per variable. The variable names must correspond to those used in object.
engine	String specifying whether to use mirt, Mplus, or TAM for estimation.
...	Not currently used. Use control instead.
link	String specifying the link function. May be either logit, or (in case of Mplus), probit.
verbose	Logical indicating whether output should be printed to the console.
control	List. The allowed elements of this list depend on the engine. Use control_mirt(), control_mplus(), or control_tam() for convenience. Note that the fit() function does not use ..., but that you can use the control_*() functions to pass additional arguments.
improper_okay	Logical indicating whether the model should also be fit if it is not a proper IR-tree model. Set this only to TRUE if you really know what you are doing.

Value

Returns a list of class irtree_fit. The first list element is the return value of either mirt::mirt(), MplusAutomation::readModels(), or TAM::tam.mml(). Further information is provided in the element spec.

Methods

The methods coef(), summary(), and print() are implemented for objects of class irtree_fit, and those wrap the respective functions of mirt, MplusAutomation, or TAM. However, glance(), tidy(), and augment() may be more helpful.

Examples

# Running these examples may take a while

data("jackson")
df1 <- jackson[1:456, paste0("C", 1:5)]
df2 <- jackson[1:456, c(paste0("C", 1:5), paste0("E", 1:5))]

irtree_create_template(df1)

# Graded Response Model ---------------------------------------------------

m1 <- "
IRT:
t BY C1@1, C2@1, C3@1, C4@1, C5@1;

Class:
GRM
"

model1 <- irtree_model(m1)

fit1 <- fit(model1, data = df1)

glance(fit1)
tidy(fit1, par_type = "difficulty")
augment(fit1)

# IR-Tree Models ----------------------------------------------------------

##### IR-tree model for 1 target trait #####

m2 <- "
Equations:
1 = (1-m)*(1-t)*e
2 = (1-m)*(1-t)*(1-e)
3 = m
4 = (1-m)*t*(1-e)
5 = (1-m)*t*e

IRT:
t BY C1@1, C2@1, C3@1, C4@1, C5@1;
e BY C1@1, C2@1, C3@1, C4@1, C5@1;
m BY C1@1, C2@1, C3@1, C4@1, C5@1;

Class:
Tree
"

model2 <- irtree_model(m2)

# See ?mirt::mirt for details on method argument
fit2 <- fit(model2, data = df1, control = control_mirt(method = "MHRM"))

##### IR-tree model for 2 target traits #####

m3 <- "
Equations:
1 = (1-m)*(1-t)*e
2 = (1-m)*(1-t)*(1-e)
3 = m
4 = (1-m)*t*(1-e)
5 = (1-m)*t*e

IRT:
t1 BY C1@1, C2@1, C3@1, C4@1, C5@1;
t2 BY E1@1, E2@1, E3@1, E4@1, E5@1;
e  BY C1@1, C2@1, C3@1, C4@1, C5@1, E1@1, E2@1, E3@1, E4@1, E5@1;
m  BY C1@1, C2@1, C3@1, C4@1, C5@1, E1@1, E2@1, E3@1, E4@1, E5@1;

Class:
Tree

Constraints:
t = t1 | t2
"

model3 <- irtree_model(m3)

fit3 <- fit(model3, data = df2, control = control_mirt(method = "MHRM"))

##### IR-tree model constrained to Steps Model #####

m4 <- "
Equations:
1 = (1-a1)
2 = a1*(1-a2)
3 = a1*a2*(1-a3)
4 = a1*a2*a3*(1-a4)
5 = a1*a2*a3*a4

IRT:
a1 BY C1@1, C2@1, C3@1, C4@1, C5@1;
a2 BY C1@1, C2@1, C3@1, C4@1, C5@1;
a3 BY C1@1, C2@1, C3@1, C4@1, C5@1;
a4 BY C1@1, C2@1, C3@1, C4@1, C5@1;

Class:
Tree

Constraints:
a1 = a2
a1 = a3
a1 = a4
"

model4 <- irtree_model(m4)

fit4 <- fit(model4, data = df1)

# Partial Credit Model ----------------------------------------------------

##### Ordinary PCM #####

m5 <- "
IRT:
t BY C1@1, C2@1, C3@1, C4@1, C5@1;

Weights:
t = c(0, 1, 2, 3, 4)

Class:
PCM
"

model5 <- irtree_model(m5)

fit5 <- fit(model5, data = df1 - 1, engine = "tam")

##### Multidimensional PCM with constraints #####

m6 <- "
IRT:
t1 BY C1@1, C2@1, C3@1, C4@1, C5@1;
t2 BY E1@1, E2@1, E3@1, E4@1, E5@1;
e  BY C1@1, C2@1, C3@1, C4@1, C5@1, E1@1, E2@1, E3@1, E4@1, E5@1;
m  BY C1@1, C2@1, C3@1, C4@1, C5@1, E1@1, E2@1, E3@1, E4@1, E5@1;

Weights:
t = c(0, 1, 2, 3, 4)
e = c(1, 0, 0, 0, 1)
m = c(0, 0, 1, 0, 0)

Class:
PCM

Constraints:
t = t1 | t2
"

model6 <- irtree_model(m6)

fit6 <- fit(model6, data = df2 - 1, engine = "tam",
            control = control_tam(control = list(snodes = 1234)))

ItemResponseTrees documentation built on July 2, 2020, 2:25 a.m.