IRTpp-multidimensional-cpp

Item Response Theory software

IRT package for multidimensional polytomous models.

Usage

Necessary to load a matrix object where dataset is allocated. This operation is performed by an input object.

matrix<char> Y;
input<char> in;
in.importData("datasets/admission_test.csv", Y);

admission_test.csv will look like:

1;0;1; ... ;0;1
0;1;0; ... ;1;1
.
.
.
1;1;1; ... ;0;0

In dichotomous case, or with positive integers in polytomous case.

Item parameters estimation

Estimations are perfomed by dichotomous::estimation class for unidimensional or multidimensional dichotomous cases, and unidimensional or multidimensional polytomous cases by polytomous::estimation.

Simple estimation
Just with dataset and dimension.

dichotomous::estimation e(Y, 7);
e.EMAlgorithm();
e.print_item_parameters();

Or more customized configurations for the estimation.
Dataset, dimension, item type [1PL, 2PL, 3PL], epsilon for EM convergence:

dichotomous::estimation e(Y, 7, 2, 0.0001);

Dataset, dimension, item type, EM epsilon, size cluster (amount of items by dimension), quadrature technique, amount of quadrature points, individuals' weights, custom initial values for item parameters:

std::vector<int> size_cluster = {20, 20, 15};
dichotomous::estimation e(Y, 7, 2, 0.001, size_cluster, "QMCEM", 2000, individual_weights,
"datasets/initial_values_admission_test.csv");

For latent traits estimation you can use expected a priori or maximum a posteriori approaches.

e.EAP(true);
e.MAP(true);
e.print_latent_traits();

Send true will estimate by individuals and false by patterns. If you have item parameters estimation you can load it and then estimate latent traits.

e.load_initial_values("datasets/admission_test_estimation.csv");
e.MAP(true);

Both, item parameters and latent traits estimation can be saved to files.

e.print_item_parameters("datasets/admission_test_estimation.csv");
e.print_latent_traits("datasets/admission_test_LTestimation.csv");

Log likelihood can be obtained calling e.log_likelihood().

std::cout << e.log_likelihood() << '\n';