irt (version 0.2.9)

classification_indices: Calculate classification accuracy and consistency

Description

Calculate classification accuracy and consistency

Usage

classification_indices(
  method = "recursive",
  ip = NULL,
  theta = NULL,
  theta_cs = NULL,
  raw_cs = NULL,
  resp = NULL,
  se = NULL,
  perf_categories = NULL,
  n_theta = 100,
  theta_lower_bound = -6,
  theta_upper_bound = 6,
  cat_labels = NULL
)

Value

A list of following elements:

category_prob

A numeric vector representing the performance category classification probabilities of each examinee.

ca

Marginal (overall) classification accuracy index

cc

Marginal (overall) classification consistency index

ind_cs_ca

Individual cut score classification accuracy indices. This value will only be calculated when there are more than one cut score.

ind_cs_ca

Individual cut score classification consistency indices. This value will only be calculated when there are more than one cut score.

Arguments

method

The method of classification accuracy and consistency calculation method. Following methods are available:

'rudner'

Rudner (2000, 2005) based classification accuracy and consistency indices.

Following values should be provided for this method: theta, se, theta_cs. Following values can optionally be provided for this method: perf_categories, cat_labels.

'guo'

Guo (2006) based classification accuracy and consistency indices.

Note that calculation times can be long for this method. The value of n_theta can be decreased to speed up the function but this will reduce the accuracy of the index.

Following values should be provided for this method: ip, resp, theta_cs and either one of theta or perf_categories. Following values can optionally be provided for this method: n_theta, theta_lower_bound, theta_upper_bound, cat_labels.

'recursive'

Lee (2010) based classification accuracy and consistency indices.

Following values should be provided for this method: ip, theta and either one of these theta_cs, raw_cs. Following values can optionally be provided for this method: perf_categories, cat_labels.

ip

An Itempool-class object. Item pool parameters can be composed of any combination of unidimensional dichotomous or polytomous items. Required for "guo" and "recursive" methods.

theta

A numeric vector representing the abilities of examinees. Required for 'rudner' and 'recursive' method. For "guo" method, this vector will be used to get performance category of each examinee if perf_categories is NULL. The default value is NULL. For method = "guo" either theta or perf_categories should be provided.

theta_cs

A sorted (ascending order) numeric vector representing the theta scale cut scores. Do not include -Inf or Inf. Required for 'rudner' and 'guo' method; required for 'recursive' if raw_cs is not provided.

raw_cs

A sorted (ascending order) numeric vector of summed-score cut score values. Do not include 0 or the maximum possible score of the test in this vector. Required for 'recursive' method if 'theta_cs' is not provided.

resp

A Response_set-class, a matrix or a data.frame object that holds responses. If matrix or a data.frame provided, they will be converted to a Response_set-class. Required for 'guo' method.

se

A numeric vector representing the standard errors of ability estimates. Required for 'rudner' method.

perf_categories

An integer vector representing the performance categories of examinees. The number 1 should represent the lowest category. For example if there are three cut scores the valid values can only be: 0, 1, 2 and 3. This vector will be used theta is NULL. The default value is NULL. Either theta or perf_categories should be provided. Can optional be provided for all methods.

n_theta

An integer representing the number of equally spaced theta points between cut scores. The default value is 100. Use larger values to increase accuracy but larger numbers will also slow the speed of calculation. Can optionally be provided for the 'guo' method.

theta_lower_bound

A number representing the lower bound for cut scores. The default value is -6. Can optionally be provided for the 'guo' method.

theta_upper_bound

A number representing the upper bound for cut scores. The default value is 6. Can optionally be provided for the 'guo' method.

cat_labels

A string vector representing the labels of the categories. The length of the vector should be one more than the length of the cut scores. The default value is NULL where the categories will be labeled as 1, 2, ..., (number of cut scores plus one). For example, if there are three cut scores category labels can be: c("Unsatisfactory", "Basic", "Mastery", "Advanced"). Can optional be provided for all methods.

Author

Emre Gonulates

References

Guo, F. (2006). Expected classification accuracy using the latent distribution. Practical Assessment, Research, and Evaluation, 11(1), 6.

Lee, W. C. (2010). Classification consistency and accuracy for complex assessments using item response theory. Journal of Educational Measurement, 47(1), 1-17.

Rudner, L. M. (2000). Computing the expected proportions of misclassified examinees. Practical Assessment, Research, and Evaluation, 7(1), 14.

Rudner, L. M. (2005). Expected classification accuracy. Practical Assessment, Research, and Evaluation, 10(1), 13.

Wyse, A. E., & Hao, S. (2012). An evaluation of item response theory classification accuracy and consistency indices. Applied Psychological Measurement, 36(7), 602-624.

Examples

ip <- generate_ip(model = sample(c("GPCM", "2PL"), 20, TRUE))
n_examinee <- 100

true_theta <- rnorm(n_examinee)
resp_set <- generate_resp_set(ip = ip, theta = true_theta, prop_missing = .2)
theta_est <- est_ability(resp = resp_set, ip = ip, method = "eap")
se <- theta_est$se
theta_est <- theta_est$est
raw_score <- est_ability(resp = resp_set, method = "sum_score")$est

# Cut score
theta_cs <- c(-1, 0, 1.5)
raw_cs <- round(rsss(ip = ip, scale_score = theta_cs))

# Rudner (2000, 2005) based indices:
classification_indices(method = "rudner", theta = theta_est, se = se,
                       theta_cs = theta_cs)

# Guo (2006) based indices:
classification_indices(method = "guo", ip = ip, resp = resp_set,
                       theta = theta_est, theta_cs = theta_cs)

# Recursive method based indices:
classification_indices(method = "recursive", ip = ip, theta = theta_est,
                       theta_cs = theta_cs)
# Use raw score cut scores with recursive method
classification_indices(method = "recursive", ip = ip, theta = theta_est,
                       raw_cs = raw_cs)