1、B3_NLS - Pred on 2011-3-5 17:12:3112345678網(wǎng)絡(luò)課堂：S主講人：A 專(zhuān)題 *博士*單電主位：中山大學(xué)嶺南學(xué)院金融系郵:9*1011121314151617181920212223242526272829303132333435363738394041424344454647484950515253545556575859606162636465666768697071727374頁(yè):/arlion:高級(jí)部分:計(jì)量分析與Sa應(yīng)用=第三講 非線性最小二乘法=*cdD:sa9adoalNet courseB3 NLS*簡(jiǎn) 介非線性模型估計(jì)方法：非線性最小二乘法

2、、最大似然法(第四講)NLS：轉(zhuǎn)化為線性模型( do B3_ro.do參考資料：展開(kāi)等),數(shù)值算法 近*Davidson, R. and J.G. MacKinnon, Eoconometric Theory and Methods, Oxford University Press, 2004, Chp6.Greene, W.H.,Econometric Prentice Hallysis,ernational, Inc, 2000, Chp10.*PartI：Sa8 用戶*簡(jiǎn)介：程序的基本架構(gòu)*m： y = B0*(1 - e(-B1*x)doedit nlexpgr.ado adopath

3、 + D:sa9adoalNet courseB3 NLSuse B3 production.dta, clear gen x = capitalnl expgr lnoutnl expgr lnout, init(B0=0, B1=2)nl expgr lnout, init(B0=3.4182, B1=6.8062)doedit nlexpgr2.ado/*設(shè)定模型的標(biāo)題*/ nl expgr2 lnout*模型的基本架構(gòu)program definever8.0if 1 = ?global S_1 參數(shù)名稱(chēng) (global. 設(shè)定參數(shù)的初始值) exitreplace 1 = .定義模型的

4、形式end執(zhí)行：nl 程序名稱(chēng) 被解釋變量*Page 1B3_NLS - Pred on 2011-3-5 17:12:31范例2：生產(chǎn)函數(shù)模型: lny = B0 + A*lnD*Lrho + (1-D)*Krho參數(shù)： B0, A, D, rho被解釋變量：lny；解釋變量：L(勞動(dòng)力投入), K(資本存量)75767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130

5、131132133134135136137138139140141142143144145146147148*doedit nl1.adonl1 lnoutput* 范例3：一般化部分調(diào)整模型. 中國(guó)上市公司資本結(jié)構(gòu)動(dòng)態(tài)調(diào)整機(jī)制研究.南方經(jīng)濟(jì),2007,(01):23-38.*,模型：Y t - Y t-1 = A(Yo t - Y t-1) 0Ar(mean), /rhta) initial(rho 1 delta 0.5)est estub, s(N r2 r2 a) b(%6.3f)*穩(wěn)健型估計(jì)nl9b lnoapital labor,/parameters(b0rhta) initia

6、l(rho 1 delta 0.5) /robustest eststore m robusttable m_full m_robust, s(N r2 r2_a) b(%6.3f) se(%6.3f)*返回值eret list擬合值和殘差*Page 3B3_NLS - Pred on 2011-3-5 17:12:31223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266nl9b ln

7、oapital labor,/parameters(b0rhta) initial(rho 1 delta 0.5)predict y_hatpredict e, resqui sum labornl9b lnoapital labor if laborr(mean),/parameters(b0rhta) initial(rho 1 delta 0.5)predict y hat2predict y hat3 if e(sle)list lnout y_hat2 y_hat3*錯(cuò)誤信息與程序的調(diào)試*初始值use nlB3 production.dta, clear9b lno parameters(b09b lno parameters(b09b lno parameters(b0apital labor,/rhta) initial(rho 1 delta 0.5)nlnlapital labor, /*不當(dāng)?shù)某跏贾禃?huì)產(chǎn)生問(wèn)題*/rhta) initial(rho 0 delta 0.5)apital labor, /rhta) initial(rho0