Heteroskedasticity. . reg wage black exper educ married tenure

Size: px

Start display at page:

Download "Heteroskedasticity. . reg wage black exper educ married tenure"

Garry Griffith
5 years ago
Views:

1 Heteroskedasticity. reg Source SS df MS Number of obs = 2, F(2, 2377) = Model Prob > F = Residual , R-squared = Adj R-squared = Total , Root MSE = deny Coef. Std. Err. t P> t [95% Conf. Interval] black diratio _cons reg wage black exper educ married tenure F(5, 929) = Model Prob > F = Residual R-squared = Adj R-squared = Total Root MSE = black exper educ married tenure _cons imtest Cameron & Trivedi's decomposition of IM-test Source chi2 df p Heteroskedasticity Skewness Kurtosis

2 Total predict yhat. predict uhat, res. gen uhat2=uhat^2. twoway (scatter yhat uhat2)(lfit yhat uhat2). twoway (scatter yhat uhat2)(lfit yhat uhat2). twoway (scatter uhat2 yhat)(lfit uhat2 yhat). reg wage educ exper black F(3, 931) = Model Prob > F = Residual R-squared = Adj R-squared = Total Root MSE = educ exper black _cons reg wage black F(1, 933) = Model Prob > F = Residual R-squared = Adj R-squared = Total Root MSE = black _cons

3 . su wage if black==0 Variable Obs Mean Std. Dev. Min Max wage gen errvar = 1 + educ/10. gen sqerv = (errvar)^(1/2). gen wagex = wage/sqerv. gen educx = educ/sqerv. gen experx = exper/sqerv. gen blackx = black/sqerv. gen marriedx = married/sqerv. reg wagex educx experx marriedx blackx F(4, 930) = Model Prob > F = Residual R-squared = Adj R-squared = Total Root MSE = wagex Coef. Std. Err. t P> t [95% Conf. Interval] educx experx marriedx blackx _cons imtest Cameron & Trivedi's decomposition of IM-test Source chi2 df p Heteroskedasticity

4 Skewness Kurtosis Total reg wage educ exper married black F(4, 930) = Model Prob > F = Residual R-squared = Adj R-squared = Total Root MSE = educ exper married black _cons imtest Cameron & Trivedi's decomposition of IM-test Source chi2 df p Heteroskedasticity Skewness Kurtosis Total reg wage educ exper married black [aweight = 1/sqrv] sqrv not found r(111);. reg wage educ exper married black [aweight = 1/sqerv] (sum of wgt is e+02) F(4, 930) = Model Prob > F =

5 Residual R-squared = Adj R-squared = Total Root MSE = educ exper married black _cons su errvar Variable Obs Mean Std. Dev. Min Max errvar reg wage educ exper married black [aweight = 1/errvar] (sum of wgt is e+02) F(4, 930) = Model Prob > F = Residual R-squared = Adj R-squared = Total Root MSE = educ exper married black _cons drop uhat-marriedx. reg wage educ exper black married F(4, 930) = Model Prob > F = Residual R-squared = Adj R-squared =

6 Total Root MSE = educ exper black married _cons predict uhat, res. gen u2 = uhat^2. predict yhat variable yhat already defined r(110);. drop yhat. predict yhat. gen y2 = yhat^2. gen lu2 = log(u2). reg lu2 yhat y F(2, 932) = Model Prob > F = Residual R-squared = Adj R-squared = Total Root MSE = lu2 Coef. Std. Err. t P> t [95% Conf. Interval] yhat y2-6.20e e e e-06 _cons predict ghat

7 . gen hhat = exp(ghat). reg wage married black educ exper [aweights = 1/hhat] aweights unknown weight type r(198);. reg wage married black educ exper [aweights = 1/hhat] aweights unknown weight type r(198);. reg wage married black educ exper [aweight = 1/hhat] (sum of wgt is e-02) F(4, 930) = Model Prob > F = Residual R-squared = Adj R-squared = Total Root MSE = married black educ exper _cons imtest imtest does not support weights r(101);. imtest, white imtest does not support weights r(101);. predict newres, r. predict newfit. gen nr2 = newres^2. reg nr2 newfit F(1, 933) = 16.25

8 Model e e+12 Prob > F = Residual e e+10 R-squared = Adj R-squared = Total e e+10 Root MSE = 3.1e+05 nr2 Coef. Std. Err. t P> t [95% Conf. Interval] newfit _cons gen wdum = wage > 958. tab wdum wdum Freq. Percent Cum Total reg wdum educ married exper black F(4, 930) = Model Prob > F = Residual R-squared = Adj R-squared = Total Root MSE = wdum Coef. Std. Err. t P> t [95% Conf. Interval] educ married exper black _cons rvfplot. drop uhat-newfit. predict yhat

9 . gen hhat = yaht*(1-yhat) yaht not found r(111);. gen hhat = yhat*(1-yhat). reg wdum educ exper black married [aweight = 1/hhat] (sum of wgt is e+03) Source SS df MS Number of obs = F(4, 925) = Model Prob > F = Residual R-squared = Adj R-squared = Total Root MSE =.4306 wdum Coef. Std. Err. t P> t [95% Conf. Interval] educ exper black married _cons reg wage black married exper educ, robust Linear regression Number of obs = 935 F(4, 930) = Prob > F = R-squared = Root MSE = Robust black married exper educ _cons

İnsan TUNALI 8 November 2018 Econ 511: Econometrics I. ASSIGNMENT 7 STATA Supplement

İnsan TUNALI 8 November 2018 Econ 511: Econometrics I ASSIGNMENT 7 STATA Supplement. use "F:\COURSES\GRADS\ECON511\SHARE\wages1.dta", clear. generate =ln(wage). scatter sch Q. Do you see a relationship