Estimators and tests: summary

Table 10-2 shows the estimators available under the Model menu in gretl's main window. The corresponding script command (if there is one available) is shown in parentheses. For details consult the command's entry in Chapter 10.

Table 10-2. Estimators

EstimatorComment
Ordinary Least Squares (ols)The workhorse estimator
Weighted Least Squares (wls)Heteroskedasticity, exclusion of selected observations
HCCM (hccm)Heteroskedasticity corrected covariance matrix
Heteroskedasticity corrected (hsk)Weighted Least Squares based on predicted error variance
Cochrane–Orcutt (corc)First-order autocorrelation
Hildreth–Lu (hilu)First-order autocorrelation
Autoregressive Estimation (ar)Higher-order autocorrelation (generalized Cochrane–Orcutt)
Vector Autoregression (var)Systems of time-series equations
Cointegration test (coint)Long-run relationships between series
Two-Stage Least Squares (tsls)Simultaneous equations
Logit (logit)Binary dependent variable (logistic distribution)
Probit (probit)Binary dependent variable (normal distribution)
Rank Correlation (spearman)Correlation with ordinal data
Pooled OLS ((pooled)OLS estimation for pooled cross-section, time series data

Table 10-3 shows the tests that are available under the Tests menu in a model window, after estimation.

Table 10-3. Tests for models

TestCorresponding command
Omit variables (F-test if OLS)omit
Add variables (F-test if OLS)add
Nonlinearity (squares)lmtest
Heteroskedasticity (White's test)lmtest
Autocorrelation up to the data frequencylmtest -o
Chow (structural break)chow
CUSUM (parameter stability)cusum
ARCH (conditional heteroskedasticity)arch
Normality of residualtestuhat
Panel diagnosticshausman

Some additional tests are available under the Variable menu in the main window: the augmented Dickey–Fuller test (command: adf) and the runs test of randomness (command: runs).