Version 1

v1.0.0

Fix deprecations (#248)

Version 0.11.0

Remove fit from docs since it doesn't yet have a docstring and restrict

docs to GLM module.

Add Reexport to REQUIRE since it is explicitly used

Change default cholpred constructor

• in v0.10.0 the GLM.cholpred constructor defaulted to a pivoted Cholesky decomposition which caused one of the tests for the Lasso package to fail.
• this version changes the default back to the unpivoted decomposition.

Allow rank-deficient LinearModel

• Allow for the model matrix in a LinearModel to be of less than full column rank.
• An additional, optional argument, allowrankdeficient, (with default value false) was added to the methods for fit(LinearModel, X, y) and to lm(X, y) for this purpose.
• The documentation now uses the Documenter.jl package.

Use DataFrames 0.11, CategoricalArrays, StatsModels

Use StatsModels v0.2.0 or later instead of the DataFrames package for the Formula, ModelFrame and ModelMatrix types.

v0.8.1: Allow for atol=1e-10 for predictor tests in `issubmodel`. (#197)

* Fix typo in `issubmodel`.

* Allow for atol=1e-10 for predictor tests in `issubmodel`.

* Add paramter atol to control tolerence.

* Typo.

* Change the test for nested models.

* Typo in atol default value.

Version 0.8.0

• Add ftest function to compare linear models with F-test.
• Evaluate inverse link and its derivative simultaneously.
• Clamp cloglog inverse link.
• Drop Julia 0.5 support, fix deprecations on Julia 0.7.

v0.7.0

Bump required Compat for parametric aliases

v0.2.0

Version for the 0.2.0 release of Julia