Recent Releases of girth
girth - Directory Refactor
Refactored directory structure to make it easier to find and update functions. The only major forward-facing change is the synthetic data creation. It is now its own module. This means imports now look like
python
from girth.synthetic import create_synthetic_irt_dichotomous
- Python
Published by eribean over 4 years ago
girth - Remove Numba Dependency
Removed Numba Dependency
No major features or bugs were fixed in this release. The dependency on Numba was removed to address some issue. This release, and others moving forward, should play nice with google colab and pyodide.
- Python
Published by eribean over 4 years ago
girth - Multidimensional Initial Guess
Added function to compute initial guess based on maximum likelihood factor analysis. Integrated it into multidimensional methods via an option
python
options = {'initial_guess': True | False}
option is only honored in multidimensional methods.
- Python
Published by eribean over 4 years ago
girth - Classical Test Statistics
Added Classical Test statistics to aid item analysis
Added functions
- polyserial_correlation
- reverse_score
- cronbach_alpha
- classicalteststatistics
- Python
Published by eribean over 4 years ago
girth - Multidimensional Ability Estimation
Added multidimensional ability estimates
- multidimensionalabilitymap
- multidimensionalabilityeap
- Python
Published by eribean over 4 years ago
girth - Added Factor Analysis Functions
Added factor analysis function to support all types of latent estimation. This will also be useful for rotation and setting the initial guess for multidimensional IRT functions.
Dropped support for Python < 3.7
Added Factor Analysis Functions
- principal component analysis
- principal axis factoring
- minimum rank factor analysis
- maximum likelihood factor analysis
Supporting functions
- procrustes rotation
- polychoric correlation
- sparsify loadings
- Python
Published by eribean over 4 years ago
girth - Multidimensional 2PL/GRM MML Estimation
Updates include multidimensional 2PL and GRM models. Estimation can be very time-consuming and adjusting the quadrature points must be decreased as the number of factors increases or else it might never finish.
Added Methods:
- multidimensionaltwoplmml
- multidimensionalgrmmml
- Python
Published by eribean over 4 years ago
girth - Bootstrap Standard Errors
Added standard errors via bootstrap. Updated README to reflect usage.
- Python
Published by eribean almost 5 years ago
girth - Missing Data Support v2
Minor update to remove a print statement.
- Python
Published by eribean almost 5 years ago
girth - Missing Data Support
Added missing data support for all methods.
- Python
Published by eribean almost 5 years ago
girth - December 2020 Update 2
Added EAP estimation for 2PL and GRM models. This is useful for small sample sizes.
- Python
Published by eribean over 5 years ago
girth - December 2020 Update
Updates to GGUM and minor changes to outputs.
- Python
Published by eribean over 5 years ago
girth - December 2020 Update
Mid release for new output signature.
Added the ability to estimate the latent distribution in: * Two parameter mml (through grm_mml) * Graded Response mml * Partial Credit mml * Unfolding mml
- Python
Published by eribean over 5 years ago
girth - December 2020 Release
- Added 3PL estimation parameters
- Code optimizations to decrease the estimation time
- Python
Published by eribean over 5 years ago
girth - March Release
Added:
- Estimate Ability Parameters for Dichotomous Models
- Missing Values work with OnePL/TwoPL Models
- Added Partial Credit Model parameter estimation
- Added Graded Unfolding Model synthetic data creation
- Included framework to run performance metrics
- Fixed a bug in Joint Maximum Likelihood estimation
- Python
Published by eribean about 6 years ago
girth - girth_v0.1.3
- Support for graded response model parameter estimation
- Support for missing values in unidimensional MML and JML functions
- Python
Published by eribean about 6 years ago