Library: | plm |
See also: | plmhett plmhetexog |
Quantlet: | plmhetmean | |
Description: | plmhetmean estimates the parameter part in partially linear heteroscedastic models, in which the variance is an unknown function of the mean. We use the replication technique to estimate the variance functions. |
Usage: | res = plmhetmean(mn,x,t,y,h) | |
Input: | ||
mn | scalar, replicate | |
x | n x p matrix, the design | |
t | n x 1 matrix, the design in [0, 1] | |
y | n x mn matrix, the response | |
h | p x 1 matrix or scalar, chosen bandwidth | |
Output: | ||
res.hbetals | p x 1 matrix, LS estimate of parameter | |
res.hbeta | p x 1 matrix, the estimate based on our method | |
res.hg0 | n x 1 matrix, estimate of nonparameter function based on res.hbetals | |
res.hg | n x 1 matrix, estimate of nonparameter function based on res.hbeta |
library("plm") randomize(100) n = 100 mn = sqrt(n) sig=0*matrix(3,3) sig[,1]=#(0.81,0.1,0.2) sig[,2]=#(0.1,2.25,0.1) sig[,3]=#(0.2,0.1,1) x =normal(n,3)*sig t =sort(uniform(n)) beta0=#(1.2, 1.3, 1.4) ; the true value ma = x*beta0+t^3 y =ma+0.01*(ma+1/(1+ma)).*normal(n,mn) h =0.25 res=plmhetmean(mn,x,t,y,h) res.hbetals res.hbeta ;;;;;;;;;;;;;;;for non-parametric part ddpt=createdisplay(1,1) datah1=t~t^3 datah2=t~res.hg0 datah3=t~res.hg part=grid(1,1,rows(t))' setmaskp(datah1,1,0,1) setmaskp(datah2,4,0,3) setmaskp(datah3,7,0,5) setmaskl(datah1,part,1,1,1) setmaskl(datah2,part,4,1,3) setmaskl(datah3,part,2,1,1) show(ddpt,1,1,datah1,datah2,datah3) setgopt(ddpt,1,1,"xlabel","T","title","Simulation comparison","ylabel","g(T) and its estimate values")
The parameter estimates, see Hua Liang and Wolfgang Haerdle" Asymptotic normality of parametric regression part in partial linear heteroscedastic regression models", DP 970033 of SFB 373.