Stata - Generalized Linear Models

New in Stata 10

Link functions
  • Identity
  • Log
  • Logit
  • Probit
  • Complementary log-log
  • Power
  • Odds power
  • Negative binomial
  • Log-log
  • Log-complement

Families

  • Gaussian (normal)
  • Inverse Gaussian
  • Bernoulli/binomial
  • Poisson
  • Negative binomial
  • Gamma

Choice of estimation method

  • Maximum likelihood
  • Iteratively reweighted least squares (IRLS)

GEE estimation for panel data

Customizable functions

  • User-defined link functions
  • User-defined variance functions
  • User-defined HAC kernels

Predicts

  • Expected value of dependent variable
  • Anscombe residual
  • Cook's distance
  • Deviance residual
  • Diagonal of hat matrix
  • Likelihood residual
  • Pearson residual
  • Response residual
  • Score residual
  • Working residual

Marginal effects

  • Marginal effects and elasticities
  • Standard errors and confidence intervals
  • Computation at means or specified covariate values
  • Computation for any predicted statistic

Choice of variance estimates and standard errors

  • Inverse Hessian
  • Outer product of the gradients (OPG)
  • Observed information matrix
  • Expected information matrix
  • Robust Huber/White/sandwich estimator
  • Robust variance with clustered/correlated data
  • Heteroskedasticity- and autocorrelation-consistent (HAC) with Newey–West, Gallant, Anderson, or user-written kernel
  • Jackknife
  • Bootstrap
Back to Capabilities Home

Back to Stata homepage
Back to Timberlake Consultants

©Timberlake Consultants Limited
Last revised:15/06/2007