* Chapter 15 - Pooling with Cross-Section Heteroskedasticity and * Cross-Section Correlation. * W.H. Greene, Econometric Analysis, Fourth Edition, 2000. SAMPLE 1 100 READ (grunfeld.shd) / NAMES * ===> Example 15.2, p. 594 * Pooling by OLS * - LM test statistic for cross-section heteroskedasticity * (formula at bottom of p. 596) * - Breusch-Pagan LM test for a diagonal error covariance matrix * (formula in Equation (15-14) p. 601). * Note that this statistic is calculated from the OLS residuals. * The textbook recommends calculation based on the residuals * obtained from the iterated estimator. POOL I F C / NCROSS=5 OLS DN STDERR=OLS_SE * Beck-Katz standard errors POOL I F C / NCROSS=5 OLS DN HETCOV STDERR=BECK * ===> Example 15.3 - results in Table 15.1, p. 598 * Allow for cross-section heteroskedasticity * - residual variances (PHI matrix) - based on OLS residuals. * - FGLS estimates POOL I F C / NCROSS=5 DN RHO=0 NOMULSIGSQ * - MLE (iterated GLS) * - Note that the residual variances reported at iteration 1 * correspond to the variances constructed from the GLS * residuals reported in Table 15.1. POOL I F C / NCROSS=5 DN RHO=0 NOMULSIGSQ ITER=20 * Alternative estimates of standard errors, Table 15.1, p. 598 OLS I F C / HETCOV STDERR=WHITE POOL I F C / NCROSS=5 OLS HETCOV CSINDEX=FIRM STDERR=CORRECT SAMPLE 1 3 FORMAT(4F15.5) PRINT OLS_SE WHITE CORRECT BECK / FORMAT * ===> Example 15.4, pp. 601-603 SAMPLE 1 100 * Allow for cross-section correlation - Table 15.2, p. 602 * - FGLS estimates POOL I F C / NCROSS=5 DN RHO=0 NOMULSIGSQ FULL STOP