> #China production function noparametric estimate
> rt<-read.table("china.txt",header=TRUE); rt
   year   realgdp  capital labor
1  1979  4016.290   4493.1 41024
2  1980  4194.429   5152.7 42361
3  1981  4460.196   5835.5 43725
4  1982  4882.096   6700.4 45295
5  1983  5322.089   7794.1 46436
6  1984  6135.583   9123.9 48197
7  1985  6923.494  10893.4 49873
8  1986  7522.190  12877.9 51282
9  1987  8195.861  14889.5 52783
10 1988  8645.281  16957.5 54334
11 1989  8247.404  18431.3 55329
12 1990  8940.758  19783.6 56740
13 1991 10088.204  21428.6 58360
14 1992 11577.152  23813.8 59432
15 1993 13525.485  27064.5 60220
16 1994 14813.392  30770.9 61470
17 1995 15927.664  34712.5 67947
18 1996 17251.361  38966.8 68850
19 1997 18478.955  43547.6 69600
20 1998 19738.047  48916.7 69957
21 1999 21046.761  54518.2 71394
22 2000 22753.226  60400.8 72085
23 2001 24936.476  67044.9 73025
24 2002 27762.468  74964.9 73740
25 2003 31091.589  84902.1 74432
26 2004 35164.634  95584.3 75200
27 2005 40237.263 108940.2 75825
> y=log((rt$realgdp/rt$labor)*10000)
> x=log((rt$capital/rt$labor)*10000)
>
>
> #package np, nonparametric and semiparametric kernel methods
> library(np)
> h1=npregbw(formula=y~x); h1

Multistart 1 of 1 |
Multistart 1 of 1 |
Multistart 1 of 1 |
Multistart 1 of 1 /
Multistart 1 of 1 |
Multistart 1 of 1 |
                  

Regression Data (27 observations, 1 variable(s)):

                       x
Bandwidth(s): 0.06893708

Regression Type: Local-Constant
Bandwidth Selection Method: Least Squares Cross-Validation
Formula: y ~ x
Bandwidth Type: Fixed
Objective Function Value: 0.001361645 (achieved on multistart 1)

Continuous Kernel Type: Second-Order Gaussian
No. Continuous Explanatory Vars.: 1

> fit1<-npreg(bws=h1,txdat=x,tydat=y)
> npplot(x,y)
错误于UseMethod("npplot", bws) :
  "npplot"没有适用于"c('double', 'numeric')"目标对象的方法
> points(x,y)
错误于plot.xy(xy.coords(x, y), type = type, ...) : 还没有调用plot.new
>
> #package KernSmooth, local linear regression
> library(KernSmooth)
> h2=dpill(x,y); h2
[1] 0.8108148
> fit2<-locpoly(x,y,bandwidth=h2)
> plot(x,y)
> lines(fit2)
> #first order derivatives
> order1<-locpoly(x,y,bandwidth=h2,drv=1)
> plot(order1$x,order1$y,type='l')
>
>
> #semiparametric regression, package np
> y1=log(rt$realgdp*100000000)
> x1=log(rt$labor*10000)
> x2=log(rt$capital*100000000)
>
> h3<-npplregbw(y1~x1|x2); h3

Multistart 1 of 1 |
Multistart 1 of 1 |
Multistart 1 of 1 |
Multistart 1 of 1 /
Multistart 1 of 1 |
Multistart 1 of 1 |
                  

Multistart 1 of 1 |
Multistart 1 of 1 |
Multistart 1 of 1 |
Multistart 1 of 1 /
Multistart 1 of 1 |
Multistart 1 of 1 |
                  

Partially Linear Model
Regression data (27 observations,
1 linear parametric variable(s) 1 nonparametric variable(s):
                   y(z)
Bandwidth(s): 0.0753333

                    x(z)
Bandwidth(s): 0.08117952

Regression Type: Local-Constant
Bandwidth Selection Method: Least Squares Cross-Validation
Formula: y1 ~ x1 | x2
Bandwidth Type: Fixed

Continuous Kernel Type: Second-Order Gaussian
No. Continuous Explanatory Vars.: 1

> #h3<-npplregbw(xdat=x1,ydat=y1,zdat=x2); h3
> fit3<-npplreg(bws=h3)
> summary(fit3)

Partially Linear Model
Regression data: 27 training points, in 2 variable(s)
With 1 linear parametric regressor(s), 1 nonparametric regressor(s)

                   y(z)
Bandwidth(s): 0.0753333

                    x(z)
Bandwidth(s): 0.08117952

                       x1
Coefficient(s): 0.5709484

Kernel Regression Estimator: Local-Constant
Bandwidth Type: Fixed

Residual standard error: 0.0001465206
R-squared: 0.999693

Continuous Kernel Type: Second-Order Gaussian
No. Continuous Explanatory Vars.: 1

> coef(fit3)
       x1
0.5709484
> npplot(bws=h3)
>