Welcome to IRTorch’s documentation!

IRTorch is a python package for analyzing response data using Item Response Theory models. The package utilizes PyTorch for parameter optimization and GPU support.

Note

The package source code available at the GitHub repository. This project is under active development.

Installation 💻

• Installation
  • Install on your own machine
  • Run from a Cloud Environment

Examples 🚀

• National mathematics example
  • Load the libraries
  • The test data
  • Construct the model
  • Fitting the model
  • Estimation of latent traits
  • Item response functions
  • Item parameters
  • Evaluate the model fit
• Swedish SAT example
  • Load the libraries
  • The test data
  • Initializing and fitting models
  • Estimation of latent traits
  • Evaluate the model fit
  • Modeling incorrect responses
  • Information

API Reference 📖

• IRT models
  • BaseIRTModel
  • OneParameterLogistic
  • TwoParameterLogistic
  • ThreeParameterLogistic
  • GeneralizedPartialCredit
  • GradedResponse
  • NestedLogit
  • NominalResponse
  • MonotoneBSpline
  • MonotoneNN
  • MonotonePolynomial
  • SurprisalSpline
  • ModelMix
• Estimation algorithms
  • BaseIRTAlgorithm
  • AE
  • JML
  • MML
  • VAE
• Model evaluation
  • Evaluator
• Plotting
  • Plotter
• Rescaling
  • Scale
  • Bit
  • Flow
  • LinkCommonItems
  • RankCDF
  • Reverse
  • Rotate
• Package configuration
  • set_verbosity()
• Datasets
  • big_five()
  • swedish_national_mathematics_1()
  • swedish_national_mathematics_2()
  • swedish_sat()
  • swedish_sat_binary()
  • swedish_sat_quantitative()
  • swedish_sat_verbal()
• Utility functions
  • cross_validation()
  • fit_multiple_models_cpu()
  • gauss_hermite()
  • get_item_categories()
  • impute_missing()
  • imv()
  • set_seed()
  • split_data()
• Torch modules
  • MonotonePolynomialModule
  • NeuralSplineFlow
  • RationalQuadraticSpline

Literature 🧾

• Literature