第1个回答 2016-01-07
estat hettest,iid
estat hettest rooms crime ldist,iid whitetst
*Baum 6.2.1 加权最小二乘回归WLS generate rooms2=rooms^2
regress lprice rooms crime ldist [aw=1/rooms2]
*---------------------------------------------------------------------- *Example 8.4 Multiplicative Heteroscedasticity. insheet using tablef6-1.csv,clear generate lnC=log(c) generate lnQ=log(q) generate lnQsq=lnQ^2 generate lnPf=log(pf)
*这里省略了企业和年份虚拟变量，并且将载重因素（LF）变量出现在方差项而非OLS regress lnC lnQ lnQsq lnPf //OLS regression *Breusch-Pagan LM test or,"bpagan LF" estat hettest lf *
predict double e,resid *robust std err
regress lnC lnQ lnQsq lnPf,vce(robust)
*FIGURE 8.2 Plot of Residuals Against Load Factor
twoway scatter e lf,xlabel(0.4(0.05)0.7) yline(0,lcolor(1))xline(0.45 0.5 0.55 0.6 0.65,lstyle(grid)) /// || function y=0.8-2*x,range(0.4 0.61) lcolor(1) lpattern(longdash) /// || function y=-0.8+2*x,range(0.4 0.61) lcolor(1)lpattern(longdash) legend(off) /// title("FIGURE 8.2 Plot of Residuals Against Load Factor") xtitle("Load Factor")
*-----------FGLS----------
generate double lne2=log(e^2)
regress lne2 lf //也可以使用非线性回归直接估计exp(z'a) predict double s2hat generate wi=exp(-s2hat)
regress lnC lnQ lnQsq lnPf [aweight=wi]
*------------------------------------------------------------------------------ *Example 8.5 Groupwise Heteroscedasticity *数据整理
quietly infile str20 COUNTRY YEAR LGASPCAR LINCOMEP LRPMG LCARPCAP /// using gasoline.dat,clear drop in 1
gen ID=ceil(_n/19) // country ID本回答被网友采纳