EViews Features
Estimation

Regression

  • Linear and nonlinear ordinary least squares (multiple regression).
  • Linear regression with PDLs on any number of independent variables.
  • Analytic derivatives for nonlinear estimation.
  • Weighted least squares.
  • White and Newey-West robust standard errors.
  • Linear quantile regression and least absolute deviations (LAD), including both Huber’s Sandwich and bootstrapping covariance calculations.
  • Stepwise regression with 7 different selection procedures available.

Instrumental Variables and GMM

  • Linear and nonlinear two-stage least squares/instrumental variables (2SLS/IV) and Generalized Method of Moments (GMM) estimation.
  • White GMM weighting for cross section data.
  • HAC GMM weighting for time series data. HAC options including prewhitening, quadratic or Bartlett kernels, and fixed, Andrews, or Newey-West bandwidth selection methods.

ARMA and ARMAX

  • Linear models with autoregressive moving average, seasonal autoregressive, and seasonal moving average errors.
  • Nonlinear models with AR and SAR specifications.
  • Estimation using the backcasting method of Box and Jenkins, or by conditional least squares.

ARCH/GARCH

  • GARCH(p,q), EGARCH, TARCH, Component GARCH, Power ARCH, Integrated GARCH.
  • The linear or nonlinear mean equation may include ARCH and ARMA terms; both the mean and variance equations allow for exogenous variables.
  • Normal, Student’s t, and Generalized Error Distributions.
  • Bollerslev-Wooldridge robust standard errors.
  • In- and out-of sample forecasts of the conditional variance and mean, and permanent components.

Limited Dependent Variable Models

  • Binary Logit, Probit, and Gompit (Extreme Value).
  • Ordered Logit, Probit, and Gompit (Extreme Value).
  • Censored and truncated models with normal, logistic, and extreme value errors (Tobit, etc.).
  • Count models with Poisson, negative binomial, and quasi-maximum likelihood (QML) specifications.
  • Huber/White robust standard errors.
  • Count models support generalized linear model or QML standard errors.
  • Hosmer-Lemeshow and Andrews Goodness-of-Fit testing for binary models.
  • Easily save results (including generalized residuals and gradients) to new EViews objects for further analysis.

Panel Data/Pooled Time Series, Cross-Sectional Data

  • Linear and nonlinear estimation with additive cross-section and period fixed or random effects.
  • Choice of quadratic unbiased estimators (QUEs) for component variances in random effects models: Swamy-Arora, Wallace-Hussain, Wansbeek-Kapteyn.
  • 2SLS/IV estimation with cross-section and period fixed or random effects.
  • Estimation with AR errors using nonlinear least squares on a transformed specification.
  • Generalized least squares, generalized 2SLS/IV estimation, GMM estimation allowing for cross-section or period heteroskedastic and correlated specifications.
  • Linear dynamic panel data estimation using first differences or orthogonal deviations with period-specific predetermined instruments (Arellano-Bond).
  • Robust standard error calculations include seven types of robust White and Panel-corrected standard errors (PCSE).
  • Testing of coefficient restrictions, omitted and redundant variables, Hausman test for correlated random effects.
  • Panel unit root tests: Levin-Lin-Chu, Breitung, Im-Pesaran-Shin, Fisher-type tests using ADF and PP tests (Maddala-Wu, Choi), Hadri.

User-Specified Maximum Likelihood

  • Use standard EViews series expressions to describe the log likelihood contributions.
  • Examples for multinomial and conditional logit, Box-Cox transformation models, disequilibrium switching models, probit models with heteroskedastic errors, nested logit, Heckman sample selection, and Weibull hazard models.

This EViews Web is Copyright of Timberlake Consultants and QMS

Last Revised:1/16/2008