[Gretl-users] Missing value handling and correlated random effects CRE model

I tried to replicate the following stata webpage example in gretl by creating cwc variables for tenure and age. After running, the number of observations are different and race_1 coefficient is very different although other coefficients are comparable. Any suggestions? https://www.stata.com/stata-news/news39-4/correlated-random-effects-models/ Model 1: Random-effects (GLS), using 28522 observations Included 4699 cross-sectional units Time-series length: minimum 1, maximum 15 Dependent variable: ln_wage Standard errors clustered by unit coefficient std. error z p-value --------------------------------------------------------- const 1.07919 0.0308898 34.94 2.07e-267 *** tm_tenure 0.0586924 0.00210393 27.90 2.93e-171 *** tm_age 0.0111681 0.00113419 9.847 7.08e-023 *** Drace_1 0.131393 0.0117410 11.19 4.52e-029 *** Drace_3 0.236570 0.0592482 3.993 6.53e-05 *** Mean dependent var 1.675133 S.D. dependent var 0.478001 Sum squared resid 5607.429 S.E. of regression 0.443427 Log-likelihood −17274.26 Akaike criterion 34558.52 Schwarz criterion 34599.81 Hannan-Quinn 34571.80 rho 0.311897 Durbin-Watson 1.017081 'Between' variance = 0.106527 'Within' variance = 0.102584 mean theta = 0.577464 corr(y,yhat)^2 = 0.13988 Joint test on named regressors - Asymptotic test statistic: Chi-square(4) = 1467.73 with p-value = 0 Breusch-Pagan test - Null hypothesis: Variance of the unit-specific error = 0 Asymptotic test statistic: Chi-square(1) = 20404.8 with p-value = 0 Data: https://www.dropbox.com/scl/fi/zd5od9r00esst52obcxtm/nlsdata.dta?rlkey=nu...