六步学会用M精编B做空间计量回归详细步骤

1、六步学会用M精编B做空间计量回归详细步公司标准化编码QQX96QT-XQQB89Q8-NQQJ6Q8-MQM9N与MATLAB链接:Excel：选项一一加载项一一COM加载项一一转到一一没有勾选项CC） M加裁项可用力FoxitReader PDF Creator COM Add-in,-Inquire耶肖Microsoft Power Map for ExcelMicrosoft Power Pivot for ExcelF1Microsoft Power View for Excel中文转耐餓项删除2. MATLAB安装目录中寻找 toolbox- ex link-点击,启用宏E: MAT

2、LAB toolbox exl ink侈収口朗KJ41 COntentS2013/12/28 2:14MATLAB COde1 KBfib excllink2013/10/9 23:48MiCrOSOft EXCel.108 KB：丫 excllink20032013/10/9 23:48MiCrOSOft EXCel -305 KBS* ExliSamp2009/11/5 19:58MiCrOSOft ExCel.72 KB4, getfu notion help2010/12/4 2:17EATLAB COde2 KB4, getfunctionlis t2010/12/4 2:17MAT

3、LAB COde3 KBMgetfunctionsign ature2012/5/4 2:59MATLAB COde4 KB%1 info2007/7/24 0:22XMLA1 KB然后，EXCel中就出现MATLAB工具二口 I A脅B愆共享野|X廊益器打X单锁卿LA辽具*（注意EXCel中的数据：）Zv4U 4.IJIJZJJJU.iaZblJ4U.IlJlJZd/4bbbdbJW14.11214.b29414.8402420.1801810.1585194.775782579.2452.129424.6839810.118179-0.03264.8548996354.14965.1294

4、34.3956830.3505260.1754084.8564535152.23950.229444.5849670.1200770.03264.5685861862.11465.429454.5664290.2295380.2408064.7582615017.33853.829464.7077270.1002870.1313914.711692483.3360.5L Sheet33 -启动 mat lab(1)点击 Start MATLAB（2） senddat. to matlab，并对变量矩阵变量进行命需（注意：选取变量为数 值，不包括各变量）-a ItalT1%I43565-231

5、| ?IWI c-如C40V.仃462T1S A 0 二二二住& fl*.W= 4 %一.8 OX% fJ爲哄A 0aw：心tKlgoeoQ14S4Z234 13 330Q24S263M-02402 XJ12SK020613 0纽03(% | B0 UO633-HKDa44S5562r1 1-:424550勒06SQilOi（切C1W2X4 556： ?toi5 1VH321曲ei-i02冋4 3.104%Tn;ICAt02沖 W2r0214M1W2J3OW202刃24 2Wh0214 72334210641QUta2闾科022M2%2:00冏0lW)C012 九4-0621 77RW1OKJ

6、W202ftl41 1CA714 356 他 AIav- | qU7A4I公3M3i$nW2124 12C72 1K353V75631399 ” B2T$U451 eH15BMnOPQR3（dckd表中数据进行命名）000（空间权重进行命名）(3)导入MATLAB中的两个矩阵变量就可以看见详细宿定人工作区:MATLAB R2Zt-rr-AL换1 ,绘图序/& UB&码&运fM士时齡證BSimlllink 嵋尸 L膵本打开圍该【作区&清除工作区清瞬令 4 Bl崗尤c4.将elhorst和jplv7两个程序文件夹复制到MATLAB安装目录的toolbox文件夹elhorst2017/1/14 19

7、:08文件夹jp I v72017/1/15 0:46 文件夹5 设置路径:布局ink邑社I応|更改MATLAB用于宣找INK/将文件舞加到路径丿 丽(EJ MATLAB toolbox 圖囹片A舍称侈改日期2017/1/1612:19 丈件夹Ajp | v7罰迅矛下载|0C3|2017/1/14 9:34丈借夹J育乐Wllte2017/1/14 9:22一 7map2017/1 门 4 9:23砧家庭毎matlab2017/1/14 9:24文険Matlab PrOgrafT | 2003-20142017/1/15 2:59:%计箕机matlabxl mhr2017/1/14 9:23益本

8、地注C)201 刀 1/14 9:25mpc2017/1/14 9:24专业(D:)nnet2017/1714 9:25一丽(E)OPC2017/1/14 9:25Osj 舸(R)ODtim2017.1/14 9:25文得夹 : CD SK%三 ifn4I川文件范jp I v7选举文件夬馭肖6 输入程序，得出结果10M TD T la夕 匸 w. 八小命令施口 1=30;N 二坯 6:W二nor三(WL):y 二A(Fj3);x二A(;j 6);XConStant=OlneS (N*T? 1): r nobs Kl=SLZe(x): j 12V 丄30；N=46；W=normw(Wl)； y=

9、A (:, 3);X=A (:, 4, 6)；XCOnStant=OneS(N*T, 1)；nobs K=size(x)；results二ols (y, xconstant x)；VnameS=StrvcatC Iogcit， ,，intercept， , Iogp，, logy)；Prt_reg (results, Vnames, 1)；Sige=*(nobs-K)/nobs)；IOglikOlS二-nobs/2*log(2*pi*sige厂1/(2*sige)* *% The (robust)LM tests developed by ElhOrStLMSarSem_Panel(resul

10、ts, W, y, rxconstant x) ； % (RObUSt)LM tests解释每一行分别表示：小该面板数据的时期数为30 （T二30）, 该面板数据有30个地区（N二30）,将空间权重矩阵标准化（W二no:TnIW （W1）,将名为A （以矩阵形式出现在MATLABA中）的变量的笫3列数扌居定义为被 解释变量y,将名为A的变量的第4、5. 6列数据定义为解释变量矩阵X,定义一个有N*T行，1列的全1矩阵，该矩阵名为:XCOnStant, （OneS即为全1 矩阵）说明解释变量矩阵X的大小：有nobs行，K列。（SiZe为描述矩阵的大小）。附录:静态面板空间计量经济学.OLS静态面

11、板编程1、普通面板编程30；N 二 46；W=normw(Wl)；y=A(:, 3);X=A( : , 4, 6)；XCOnStant二OneS(N*T, 1)；nobs K=size(x)；results二ols (y, xconstant x)；logy)；% (RObUSt) LMVnameS=StrvcatC Iogcit， , intercept， , Togp，,Prt_reg (results, Vnames, 1)；Sige二*(nobs-K)/nobs)；IOglikOIS二-nobs/2*log(2%i*sige厂1/ (2*sige)* *% The (robust)LM

12、 tests developed by ElhOrStLMSarSem_PaneI (results, W, y, FxconstalTt x) t ests2、空间固定 OLS (SPatial-fixed effects)N=46；W=normw(Wl)；y=A( : ,3);x=A(:, 4, 6)；XCOnStantpnes (*T, 1)；nobs K=size(x)；model=l；ywith, XWith, meanny, meannx, meanty, meantx=demean (y, x, N,T, mo del)；resuItS=zoIs (ywith, XWith)；V

13、nameS=Strvcat C IOgCit， , logp， , logy，)； % ShOUld be Changed if X is Changedp*Tt_reg (results, VnameS)；sfe=meanny-meannx*； % including the COnStant termyme = y - mean(y)；et二ones (T, 1)；error=y-kron (et, Sfe)-X*；rsqrl = error *er：T0r；rsqr2 = yme*yme；FE_rsqr2 =- rsqrl/rsqr2 % rsquared including fixed

14、effectsSige=*(nobs-K)/nobs)；IoglikSfe二-nobs/2*log(2*pi*sige)T/(2*sige)* *LMSarSeln_Pane 1 (results, W, ywith, xwith) ； % (RObUSt) LM tests3、时期固定 OLS (time-period fixed effects)T=30；N=46；W=normw(Wl)；y=A(:, 3)；x=A (:, 4,6)；XCOnStant二OneS(N*T, 1)；nobs K二size (x)；model=2；ywith, xwith, meanny, meannx, me

15、anty, meantx =demean (y, x, N,T, model)；results=ols(ywith, XWith)；VnameS=StrvcatC IOgCit，, logp， , logyO ； % ShOUId be Changed if X is Changed30；Prt_reg(results, VnameS)；tfe=meanty-meantx*； % InCIUding the COnStant termyme 二 y - mean(y)；en=ones (N, 1)；error=ykron (tfe, en)-X*；rsqrl = error1 *ex t or

16、；rsqr2 = yme, *yme；FE_rsqr2 =- rsqrl/rsqr2 % rsquared including fixedeffectsSige=*(nobs-K)/nobs)；IOgliktfe二-Hobs/2*log(2*pi*sige)T/(2*sige)* *LMSarSenI_PaneI (results, W, pvith, XWith) ； % (RObUSt) LM tests4、空间与时间双固定模型T=30；N=46；W-normw(Wl)；y二A(：,3)；x=A(:, 4, 6)；XCOnStant二OneS (N*T, 1)；nobs K二size(x)

17、； model二3；ywith, xwith, meanny, meannx, meanty, meantx=demean(y, x, N, T, model)；results二ols(ywith, XWith)；VnameS=Strvcat Iogcit，, logp，,，logy，)； % ShOUld be Changed if X is ChangedPrt_reg(results, VnameS) en=ones (N, 1)；et二ones (T, 1)；intercept=mean(y)-mean(x) *；sfe=meanny-meannx*(en, intercept)；tf

18、e=meanty-meantx*(et, intercept)；yme 二 y - mean(y)；ent=ones (N*T, 1)；error=y-kron (tfe, en)-kron(et, Sfe)-X* (ent, intercept)；rsqrl = error，terror；rsqr2 = yme*ynie；FE_rsqr2 =- rsqrl/rsqr2 % rsquared InCIUding fixedeffectsSige=*(nobs-K)/nobs)；IOglikstfe二-Hobs/2*log(2*pi*sige)-1/ (2*sige)* *LMSarSeln_P

19、ane 1 (results, W, ywith, xwith) ； % (RObUSt) LM tests二、静态面板SAR模型1、无固定效应(Xo fixed effects )T=30;N=46;W=normw(Wl);y二A(:, 3)；X=A(:, 4,6)；for t=l:Ttl=(t-l)*N+ljt2=t*N; wx(tl:t2, :)=W*x(tl: t2,:);endXCOnStant=OneS(N*T, 1);nobs Kl=SiZe(X);=0；=0；=0；resul ts=sar panel-FE(j Exconstant x, W, T, info); VnameS

20、=St rVCat( IOgCit,- intercept, logp, IOgy ); Prt SPneW( resul ts, Vnames, 1)% Print OUt effects estimatesSPat_model=0;directindirecteffeets_estimates(results, W, spat model);Panel effects sar (resul ts, Vnanles, W)；2、空间固定效应(SPalial fixed effects)T=30;N=46;W=normw(Wl);y=A(:, 3);X二A(:, 4,6)；for t=l:Tt

21、1=(t-1)*N+1;t2=t*N;WX(tl:t2, :)=W*x(tl:t2,:);endXCOnStant=OneS(N*T, 1);nobs Kl=SiZe(X);=0；=1；=0；resultS=Sar-Panel-FE(y, x, W, T, info);VnameS=St rVcat (* logci t， logp J logy);Prt-SPneW(results, Vnames, 1)% Print OUt effects estimates spatmode 1=0;di rect_indi rect_ef feets_est imates(r esul ts, W,

22、spat model); Pane I- effectssar(results, Vnames, W)； 3时点固定效应（Time PeriOdfixed effects)T=30;N=46;W=normw(Wl)；y=A(:, 3);X=A (:, 4, 6)；for t=l:Tt1=(t-1)*N+1;t2=t*N;wx(11:12, :)=W*x(11:t2,:);endXCOnS t an t =0neS(N*T, 1);nobs Kl=SiZe(X);=0; % requi red for exact results=2;=0; % DO not Print intercept an

23、d fixed effects; USe =1 to lurn On results=sarpanel-FE(y, x, W, T, info);VnanleS=St rVcat (* logci t, Iogp, logy);Prt-SPneW( resul ts, VnameS：!)% Print OUt effects estimatesspat mode 1=0;di rect_indi rect_ef feets est imates(resul ts, W, spat-model);Panel effeets sar(results, VnameSr W)；4、固父攵应(SPati

24、a 1 and time PeriOd fixed effects )T=30;N=46;W=normw(Wl)；y=A(:, 3);X=A (:, 4, 6)；for t=l:Tt1=(t-1)*N+1;t2=t*N;wx(11: t2, :)=W*x(11:t2,:);endXCOnS t an t=0neS(N*T, 1);nobs Kl=SiZe(X);=0; % required for exact results=3;=0; % DO not Print intercept and fixed effects; USe =1 to lun On resultS=Sar-Panel-

25、FE(y, x, W, T, info);VnameS=St rVcat (* logci t , logp*, logy);Prt-SPneW(results, Vnames, 1)% Print 0l4 effects est imates spatmodel=0; di rect_indi rect_effects_est imates(results, W, spatmode 1); Panel effects-sar (resul ts, Vnan t es, W)；三、静态面板SDM模型1无固定效应(NO fixed effects )T二 30;N=46;W=normw(Wl);

26、y=A(:, 3)；X=A (:, 4, 6)；for t=l:Ttl=(t-l)*N+ljt2=t*N;WX(tl:t2, :)=W*x(tl:t2,:);endXCOnStant=OneS(N*T, 1);nobs Kl=SiZe(X);=0；=0；=0；resul tS=Sar-Panel-FE(y, Exconstant X vx, W, T, info);VnameS=StrVCatC IOgCit interceptA, logp-, log , W*logp ,1 W* 1 ogy1;Prt-SPneW(results, Vnames, 1)% Print OUt effects

27、 estimatesspat mode 1=1;di rect_indi rect_ef feets_est imates(results, W, spat model);Panel effectS-Sdm(results, VnameSf W)；2 空间固定效应(SPalial fixed effects )T=30;N=46;W=normw(Wl)；y=A(:, 3);X=A (:, 4, 6)；for t=l:T11=(t-1)*N+1;t2=t*N;WX(tl:t2, :)=W*x(tl:t2,:);endXC0nStant=0neS(N*T, 1);nobs Kl=SiZe(X);=

28、0; % required for exact results=1；=0; % DO not Print intercept and fixed effects; USe =1 to turn On results=sar-panel-FE(y, x wxf W T, info);VnameS=StrVCat ( IOgCit/ logp*, logy1 W*logp,1 W*logy);Prt-SPneW(results, Vnames, 1)% Print OUt effects estimates spat model=l;di rect_indi rect_ef feets_est i

29、mates(results, W, spat model);Panel-effectS-Sdln(results, Vnanles, W);3 时点固定效应(Time Period fixed effects)T=30;N=46;W=normw(Wl);y=A(:, 3)；x二A(:, 4,6)；for t=l:Ttl=(t-l)*N+ljt2=t*N;WX(tl:t2, :)=W*x(tl: t2,:);endXCOnStant=OneS(N*T, 1);nobs Kl=SiZe(X);=0; % required for exact results=2;=0; % DO not Print

30、 intercept and fixed effeets; l.Se =1 to turn On % NeW routines to Calculate effects estimates results=sarpanel-FE (jx wx, W, T, info);VnameS=St rVCat ( IOgCi f logp*, IOgy, W*logp*, W*1ogy);% Print OUt Coefficient estimatesPrt-SPneW(results, Vnames, 1)% Print OUt effects estimatesSPat_model=l;di re

31、ct_indi rect_ef feets_est imates(results, W, spat model);PaneleffectS-Sdm(results, Vnanles, W)4、j又固三父攵应(SPatia 1 and time PeriOd fixed effects)T=30;N=46;W=normw(Wl)；y=A(:, 3);X=A (:, 4, 6)；for t=l:Ttl二(tl)*N+ljt2=t*N;WX(tl:t2, :)=W*x(tl:t2,:);endXC0nStant=0neS(N*T, 1); nobs Kl=SiZe(X);=0；=0; % requi

32、red for exact results=3;=0; b DO not Print intercept and fixed effects; USe =1 to turn On results=sar-panel FE(y, xvx, W, T, info);VnameS=St rVCat ( logci t r logp:;:, * logy； W*logp-, * WMogy);Prt-SPneW(results, Vnames, 1)% Print OUt effects estimatesspat model=l;di rectindi rectef feet sest i mate

33、s (results, W, spat model);PaneI effectS-Sdm(results, VnanleSi W)Wald test SPatial lag% Wald test for SPatial DUrbin model against SPa t i a I lag mode) b t emp=;VarC0V=;Rafg=ZerOS(KJ 2*K+2);for k=l:KRafg(k, K+k)=l; % R(l,3)=0 and R(2, 4)=0;endWald-SPatial-lag=(Rafg*btemp) *inV(Rafg*varcov*Rafg )*Ra

34、fg*btempPrOb-SPatiala$I-ChiS-Cdf (Wald-SPatial-lag, K)Wald test SPatial error% Wald test SPatiaI DUrbin model against SPatial error model R=ZerOS(K, 1); for k=l:KR(k)=btemp(2*K-J-l)*btemp(k)+btemp(K+k)J % k Changed in 1, 7/12/2010 %R(I)=btemp(5)*btemp(1)+btemp(3);% R(2)=btemp(5)*btemp(2)+btemp(4);en

35、dRafg=Zeros(K, 2*K+2);for k=l:KRafg (k,k) =btemp(2*K-rl); % k Changed in I 7 12/2010Rafg(k, K+k) =1;Rafg(k, 2*K+1)=btemp(k);% Rafg(l, l)=btemp(5);Rafg(l, 3)=1;Rafg(l, 5)=btemp(1);% Rafg(2, 2)=btemp(5);Rafg(2, 4)=1;Rafg(2, 5)=btemp(2);endWald-SPat ial error=R *inv(Rafg*varcov*Rsf 才)*RPrOb-SPat ial er

36、ror=l -ChiS-Cdf (Wald-spat ial_error, K)LR test SPatial lagresul tssar=Sar-PaneI-FE(y, x, W, T, info);LR-SPat ial_lag=-2* (LR-SPatial lag, K)LR test SPatial errorresul tssem=Senl Panel FE(y, x, W, T, info);LR-SPat ial_error=-2* (LR-SPat i alerror, K)5、空间随机效应与时点固定效应模型T二 30;N 二 46;W=normw(Wl);y=A(:, 3

37、);X=A (:, 4, 6):for t=l:T11=(t-1)*N+1;t2=t*N;WX(tl:t2, :)=W*x(tl:t2,:);endXCOnStant=()neS(N*T, 1);nobs Kl=SiZe(X);5wi th, XWithJ meanny, meannx, meanty, meantx=demean (y, x WX1, N, T, 2) J % 2=t ime dummies =1 ；results=sar panel RE (ywith, xwith, W, T, info);Prt-SPneW( resul ts, VnameS：l)spat mode 1

38、=1;di rec t_i nd i rec t_ef feet sest i mates (results, W, spat model);PaneI effectS-Sdln(results, Vnames, W)Wald test SPatial lagbtemp=(1:2*K+2);VarCOniz2*G2, 1:2*K=2);Rafg=Zeros(K, 2*K+2);for k=l:KRafg(k, K-k)=l; % R(l,3)=0 and R(2, 4)=0;endWald-SPatiaI Iag=(Rafg*btemp)1 *inV(Rdfg*varcov*Rafg, )*R

39、afg*btempPrOb-SPat i a LI ag= I-ChiS-Cdf (Wald-SPat ial-lagi K)Wald test SPatial errorR=ZerOS(K)l);for k=l:KR(k)=btemp(2*K-rl)*btemp(k)+btemp(K+k) J ； k Changed in 1, %7/12/2010R(I)=btemp(5)*btemp(l)+btemp(3);% R( 2 )=bt emp( 5) *bt emp( 2 )+bt emp(4);endRafg=Zeros(K, 2*K+2);for k=l:KRafg(k, k) =bte

40、mp(2*K-l); % k Changed in 1, 7 12/2010Rafg(k, K-k) =1;Rafg (k, 2*K-rI)=btemp(k);： Rafg(b l)=btemp(5);Rafg(l, 3)=1 ;Rafg(l, 5)=btemp( 1);-： Rafg (2, 2)=btemp(5); Rafg (2, 4)=1; Rafg (2, 5)=btemp(2);endWa 1 d-SPat ia 1 -error=R *inv(Rafg*varcov*Rafg*)*RPrOb-SPat ial error= I-ChiS-Cdf (Waid-SPat ial er

41、ror, K)LR test SPatial lagresul tssar=Sar Panel RE(A；i th, XWi th(: J 1:K), W, T, info); LR-SPat ial_lag=-2* (LR-SPat ial-lag,K)LR test SPatial errorresul tsSem=SenI-Panel RE(Ari th, XWith(:, 1:K)i W, Tr info);LR-SPatial_error=-2* (LR-SPat ial error, K)四、静态面板SEM模型1 无固定效应(NO fixed effects )T=30;N=46;

42、W=normw(Wl)；y=A(:, 3);X=A (:, 4, 6)；for t=l:Ttl=(t-l)*N+ljt2=t*N;wx(11: t2, :)=W*x(11:t2,:);endXCOnStant=OneS(N*T, 1);nobs Kl=SiZe(X);=O;=O；=O；resul ts=sem panel-FE(y, Exconstant x, W, T, info);VnameS=StrVCat( IOgCi t intercept, logpJ IOgy);Prt-SPneW(results, Vnames, 1)% Print OUt effects estimatesS

43、Pat_model=0;directindirecteffeets_estimates(results, W, spat model);Panel.effects.sar (resul ts, Vnanles, W)；2、空间固定效应(SPatial fixed effects)T=30;N=46;W=normw(Wl)；y=A(:, 3);X=A(:, 4,6)；for t=l:Tt1=(t-1)*N+1;t2=t*N;WX(tl:t2, :)=W*x(tl:t2,:);endXCOnStant=OneS(N*T, 1);nobs Kl=SiZe(X);=O;=1；=O；resul ts=S

44、eIn-Panel FE (y, x, W, T, info);VnameS=St rVcat (* logci t , logp J 1 1ogy);Prt-SPneW( resul ts, Vnames, 1)% Print OUt effects est imatesspat mode 1=0;di rectindi rectef feets_est imates(results, W, spat-model); Pane Ieffects.sar(results, VnameSf W);3时点固定效应(Time Periodfixed effects)T=30;N=46;W=normw

45、(Wl)；y=A(:, 3)；x二A(:, 4,6)；for t=l:Ttl二(tl)*N+ljt2=t*N;wx(t1:t2, :)=W*x(t1:t2,:);endXCOnStant=OneS(N*T, 1);nobs Kl=SiZe(X);=0; % requi red for exact results=2;=0; % DO not Prini intercept and fixed effects; USe =1 to turn On results=sempanel-FE (y, x, W, T, info);VnanleS=St rVcat (* logci t1 / logp

46、J logy);Prt-SPneW(results, VnanleS. 1)% Print OUt effects estimatesspat mode 1=0;di rect_indi rect_ef feets_est imates(results, W, spat-model);Panel effects sar (resul ts, Vnanles, W)；4、固父攵应 (SPatia 1 and time PeriOd fixed effects)T=30;N=46;W=normw(Wl);y=A(:, 3);X二A(:, 4,6)；for t=l:Ttl二(tl)*N+ljt2=t

47、*N;WX (tl:t2, :)=W*x(tl:t2,:);endXCOnS t an t=0neS(N*T, 1);nobs Kl=SiZe(X);=0; % required for exact results=3;=0; % DO not Print intercept and fixed effects; USe =1 to turn On results=sempanel-FE (y, x, W,T, info);VnameS=St rVcat (* logci t, logp: logy);Prt-SPneW(results, Vnames, 1)% Print OUt effec

48、ts estimatesspat mode 1=0;di rect_indi rect_ef feets_est imates(results, W, spat-model);Panel effeets-sar(results, Vnames, W)；五、静态面板SDEM模型1 无固定效应(o fixed effects )T=30;N=46;W=normw(Wl);y=A(:, 3);X=A (:, 4, 6): for t=l:Ttl=(t-l)*N+l;t2=t*X; wx(t1:t2, :)=W*x(t1:t2,:); endXCOnStant=OneS(N*T, 1);nobs Kl

49、=SiZe(X):=0；=0；=0；resul ts=sem panel FE(y, Exconstant X vxj, W, T, info);VnameS=StrVCat(1 IOgCit , intercept / logp*, IOgy ,W*logp W*logy);Prt-SPneW(results, Vnames, 1)% Print OUt effects estimatesspat moclel=l;di rec t_i nd i rec t_ef feet sest i mates (results, W, spat-model);PaneIeffeetSSdm(results, Vnanles, W)；2、空间固定效应(SPatial fixed effects)T=30;N=46;W=normw(Wl);y=A(:, 3);X=A (:, 4,6)；for t=l:T11=(