GLM.jl Versions Save

GLM v1.9.0

Diff since v1.8.2

Merged pull requests:

• QR decomposition method in GLM. (#507) (@mousum-github)
• Bump actions/cache from 1 to 3 (#524) (@dependabot[bot])
• Bump codecov/codecov-action from 1 to 3 (#525) (@dependabot[bot])
• Bump actions/checkout from 2 to 3 (#526) (@dependabot[bot])
• bump compat for StatsBase 0.34 (#529) (@palday)
• Set minimum Julia version to 1.6 (#514) (#530) (@palday)
• CompatHelper: bump compat for StatsModels to 0.7, (keep existing compat) (#532) (@github-actions[bot])
• CompatHelper: bump compat for StatsBase to 0.34, (keep existing compat) (#533) (@github-actions[bot])
• Remove installbeta! and make linear model fit insensitive to beta0 state (#535) (@andreasnoack)
• set version to 2.0.0-DEV (#536) (@palday)
• [v1] Allow documentation building on push (#542) (@ararslan)
• Bump actions/checkout from 3 to 4 (#547) (@dependabot[bot])
• [G]VIF (#548) (@palday)
• backport VIF to 1.x release (#549) (@palday)

Closed issues:

• Add VIF? (#428)
• Feature request: Support ForwardDiff automatic differentiation (#493)
• Feature request: Support non-integer dependent variable for Poisson regression (#494)
• I propose to consider adding precompilation to GLM.jl (#513)
• Feature request: Support for reuse of pinv computation for multiple Y values in linear models (#534)
• infinite uninterruptible hang on one illegal input to ftest (#538)
• Feature Request: Better formats for the GLM outputs (#543)

GLM v1.8.2

Diff since v1.8.1

Closed issues:

• GLM fails when Matrix contains a single column (#515)
• Method 1 and 2 produce different results for the same data (#516)

Merged pull requests:

• Get rid of TableRegressionModel by storing formula directly in model (#339) (@nalimilan)
• fix docstring of glm (#506) (@heliosdrm)
• add compat for statsmodels 0.7 (#512) (@kleinschmidt)
• Set minimum Julia version to 1.6 (#514) (@ararslan)

GLM v1.8.1

Diff since v1.8.0

Closed issues:

• easy printing of fstat (#400)
• prediction fails on custom vectors (#442)
• ftest and -0.0 (#461)
• r2 for GLM (#475)
• Link doesn't behave like an abstract type (#486)
• lrtest gives incorrect results (#490)
• Prediction with confidence intervals not yet implemented for weighted regression (#495)
• Int conversion in loglik_obs(::Binomial) (#503)

Merged pull requests:

• Implement nulldeviance and nullloglikelihood for GLMs (#479) (@nalimilan)
• Readability improvement (#482) (@mousum-github)
• Implementation of dropcollinear feature in GeneralizedLinearModel (#488) (@mousum-github)
• Docs: Swap dependent and independent (#497) (@BioTurboNick)
• ditch negative zero in ftest output (#501) (@palday)
• autocancel stale CI runs (#502) (@palday)
• safer int for binomial loglik (#504) (@palday)

GLM v1.8.0

Diff since v1.7.0

Closed issues:

• Canonical link for negative binomial distribution (#476)

Merged pull requests:

• Update the perf script. (#433) (@dmbates)
• Adding PowerLink link function in GLM (#474) (@ayushpatnaikgit)
• Fix mention of canonical link for NegativeBinomial (#478) (@nalimilan)
• Adding geometric distribution for GLM.jl (#480) (@ayushpatnaikgit)
• CompatHelper: bump compat for "StatsFuns" to "1.0" (#484) (@github-actions[bot])

GLM v1.7.0

Diff since v1.6.2

Closed issues:

• Linear models failing to display with ambiguous datatypes with unhelpful error messages (#456)
• DataFrames dependency error (#472)

Merged pull requests:

• Fix coeftable for saturated linear models (#458) (@nalimilan)
• fix predict docstring and remove trailing dim (#467) (@piever)
• Fixed spelling (#473) (@harsharora21)

GLM v1.6.2

Diff since v1.6.1

Closed issues:

• GLM.lm(X, y) fails when X is a SubArray (#213)
• Stackoverflow with views (#444)
• Serialising GLM models. (#465)
• Gamma shape and scale seemingly swapped - where am I going wrong? (#468)
• cholpred + @view results in MethodError (#470)

Merged pull requests:

• Improved support for views (#471) (@palday)

GLM v1.6.1

Diff since v1.6.0

Merged pull requests:

• Allow for sparse arrays and views (#446) (@andreasnoack)

GLM v1.6.0

Diff since v1.5.1

Closed issues:

• License terms incomplete? (#441)
• Difference between passing data directly or via formula for glm (#450)
• project.toml version and released versions appear to be out of sync (#463)

Merged pull requests:

• Implement fstatistic (#424) (@OndrejSlamecka)
• Add residuals convenience function to docs (#445) (@ym-han)
• Update license to indicate GPL dependency. (#447) (@rcurtin)
• Add ftest to the manual (#451) (@nalimilan)
• Fix manual deployment (#452) (@nalimilan)
• CompatHelper: bump compat for "SpecialFunctions" to "2.0" (#457) (@github-actions[bot])
• compat fixes for Julia v1.8 (#459) (@dkarrasch)
• Update doctests for pivot (#460) (@palday)
• use StatsModels.isnested instead of defining new issubmodel func (#462) (@kleinschmidt)
• change version back to 1.6.0 since 1.6.0 was never released (#464) (@kleinschmidt)

GLM v1.5.1

Diff since v1.5.0

Merged pull requests:

• use StatsModels.formula (#440) (@palday)

GLM v1.5.0

Diff since v1.4.2

Closed issues:

• buggy deprecation for :confint interval (#408)
• canonicallink(NegativeBinomial()) doesn't give LogLink? (#418)
• predict() gives wrong results when dropcollinear = true (#431)

Merged pull requests:

• Improve docs (#382) (@nalimilan)
• Fix deprecation of predict(..., interval=:confint) (#412) (@nalimilan)
• Cooks Distance tests (#415) (@ericqu)
• Re-add precision about NegativeBinomialLink to manual (#419) (@nalimilan)
• Add ftest to manual (#422) (@OndrejSlamecka)
• CompatHelper: bump compat for "Distributions" to "0.25" (#429) (@github-actions[bot])
• Use StableRNG for testing (#434) (@andreasnoack)
• Fix more predict confidence intervals when dropcollinear=true (#436) (@nalimilan)
• Fix GitHub Actions arch (#437) (@nalimilan)
• Throw an error in predict with intervals for rank-deficient models (#438) (@nalimilan)
• Bump version to 1.5.0 (#439) (@nalimilan)