Add Support for Additional Model Classes

[Thomas J. Leeper]

This is a list of requested model classes to potentially support in prediction. It is sorted by completion status of model support, and then by function name. In order to support a model class, the package needs to provide a predict() method. If this is not provided, it is not (presently) supportable.

Supported

• arm::bayesglm() (inherits from "glm")
• aod::betabin(), aod::negbin(), aod::quasibin(), aod::quasipois(),
• betareg::betareg() (class "betareg"), with predict() method (but no se.fit arg)
• biglm::biglm() and biglm::bigglm()
• brglm::brglm() (inherits from "glm")
• mda::bruto()
• ordinal::clm()
• crch censored/truncated regression methods (#4)
• earth::earth()
• mda::fda()
• gam::gam()
• kernlab::gausspr()
• gee::gee()
• stats::glm()
• glmnet::glmnet()
• glmx::glmx() (class "glmx")
• MASS::glm.nb() (inherits from "glm")
• glmx::hetglm() (class "hetglm")
• pscl::hurdle()
• AER::ivreg()
• caret::knnreg()
• kernlab::kqr()
• kernlab::ksvm()
• MASS::lda()
• stats::lm()
• lme4 (lmer(), glmer(), ...) (but lme4:::predict.merMod() has no se.fit argument)
• nlme
  • nlme:::predict.lme() has no se.fit or type arguments
  • nlme:::predict.gls() has no se.fit argument.
• stats::loess() (predict() method uses se rather than se.fit)
• MASS::lqs()
• mda::mars()
• MASS::mca()
• mclogit::mclogit() (class "mclogit"), which is basically "glm"
• mda::mda()
• mlogit::mlogit()
• mnlogit::mnlogit()
• MNP::mnp()
• nnet::multinom() or any nnet model generally
• e1071::naiveBayes()
• nls (requires setting model = TRUE in original call; and then getting term(object[["model"]]))
• MASS::polr() (has predict() method with type = c("class", "probs"))
  • arm::bayespolr() (covered by general "polr" method)
• mda::polyreg()
• caret::predict()
• MASS::qda()
• quantreg::rq()
• rpart::rpart()
• sampleSelection::selection() (class "select")
• speedglm::speedglm() and speedglm::speedlm()
• survival::survreg()
  • coxph class has different type values for predict(): c("lp", "risk", "expected")
• survey::svyglm()
• AER::tobit()
• truncreg::truncreg()
• pscl::zeroinfl()

Potentially supportable

• bamlss::bamlss()
• dynlm::dynlm()
• gnm::gnm()
• LARF::larf()
• bbmle::mle2()
• mpt::mpt()
• np (non-parametric regression)
• rms
• "stanreg" objects from rstanarm
• VGAM::vglm()
• mgcv::bam()
• mgcv::gam()
• gamlss::gamlss()
• mgcv::gamm()
• VGAM::vgam()
• spdep
• sphet::spreg()
• splm::spml()
• splm::spgm()

Not currently supportable

• censReg (has a margEff() generic but no predict() method)
• ordinal::clmm() (no predict() method)
• lfe::felm() (no predict() method)
• dispmod: lm.disp(), glm.binomial.disp(), glm.poisson.disp() (no predict() method)
• geepack: geeglm(), geese(), ordgee() (no predict() methods)
• ghyp (ORPHANED)
• gmm::gmm() (no predict() method)
• gmnl::gmnl() (no predict() method)
• GeneralizedHyperbolic::hyperblm() (no predict() method)
• ivprobit::ivprobit() (no predict() method)
• class::knn() (does not construct a model object)
• sem::tsls() (no predict() method)
• plm::plm() (non-exported predict method)
• plm::pglm()
• survey::svyolr() (no predict() method)

(Note: this is migrated from: leeper/margins#3)

[bgall]

@leeper I'd add the other survey package functions. While many of them are based on survey::svyglm() which are class glm, lm, and svyglm, the survey::svyolr() function is NOT from MASS::polr() and is of class svyolr.

[Kevin Tappe]

For models produced by plm::plm(), there is a predict method available since plm version 2.6-2.