选项包括

其中

familyname选项包括 gaussian | igaussian | binomial [varname|#] | poisson | nbinomial [#] | gamma，即核函数类型

linkname 选项包括如下内容：identify | log | logit | probit | cloglog | opower # | power # | nbinomial

test：要求测试带宽的重要性。这测试了gwr模型对数据的描述是否明显优于 全局的回归模型。

sample(#)指定在带宽校准过程中使用的观测值百分比，默认为100%。这是特别对于大型数据集很有用，可以减少校准带宽所需的时间。如果指定了该选项，将随机抽取#%的观测数据并用于校准过程。

bandwidth(#)允许用户输入带宽值，并减少gwr运行所需的时间。

nolog抑制带宽优化迭代的显示。

iterate(#)指定在估计带宽时允许的最大迭代次数。默认值为50。

save (filename)创建一个Stata数据文件，其中包含从计算gwr的每个点估算的参数。

outfile(filename)创建文本文件filename。

replace表示save和/或outfile指定的文件可以 被覆盖。它也适用于mcsave选项。

reps(#)指定要执行的蒙特卡罗模拟的数量。默认值为1000。

操作案例：

结果为：

.gwr cars class unemp, east(easting) north(northing) testGlobalModel

Source| SS df MS Number of obs = 120-------------+----------------------------------F(2, 117) = 287.17Model| 4.51965851 2 2.25982925 Prob > F = 0.0000Residual| .920700696 117 .007869237 R-squared = 0.8308-------------+----------------------------------Adj R-squared = 0.8279Total| 5.4403592 119 .045717304 Root MSE = .08871

------------------------------------------------------------------------------cars| Coef. Std. Err. t P>|t| [95% Conf. Interval]-------------+----------------------------------------------------------------class| .0188073 .0033449 5.62 0.000 .0121829 .0254316unemp| -.0182798 .0011238 -16.27 0.000 -.0205054 -.0160543_cons| .8847704 .0288569 30.66 0.000 .8276208 .94192------------------------------------------------------------------------------

三、Stata基本操作2

spregxt是一个完整的Stata模块工具包 ，可以估计 不同类型的空间面板计量经济回归模型。

例如(SAR - SEM - SDM - SAC - SPGMM - GS2SLS - GS2SLSAR)回归。

spregxt可以估算以下模型: 1-自相关(序列相关)回归模型。2-干扰项异方差回归模型。3-扰动项下的非正态回归模型，以及自变量多重共线性回归模型。

事实上，spregxt估计了许多空间和非空间模型，甚至OLS回归:

spregxt varlist , nc(#) model(ols) run(ols)不选择:wmfile， spregxt的回归结果与OLS相同

运行spregxt varlist , nc(#) model(ols) run(xtbe, xtfe, xtre, xtmle, xtpa, xtgls, ...)不选择:wmfile，spregxt的结果与xtreg和更多的XT命令的面板相同。

所有诊断测试对所有模型都有效。

spregxt估计连续和截断因变量模型tobit。

model(sar, sem, sdm, sac, mstar,mstard, spgmm)处理连续或截断相关的数据 变量。如果depvar有缺失值或下限，那么在这种情况下， model(sar, sem, sdm, sac, mstar,mstard, spgmm)将会通过xttobit模型拟合空间面板模型，因此spregxt可以解决 存在于多种数据中缺失值的问题。 否则，在连续数据的情况下，将使用正态估计。

spregxt语法格式为：

* xtabond spregxt y x1 x2 , nc(7) wmfile(SPWxt) model(gwr) run(xtabond) inst(x1 x2) * xtdhp spregxt y x1 x2 , nc(7) wmfile(SPWxt) model(gwr) run(xtdhp) re * xtdpd spregxt y x1 x2 , nc(7) wmfile(SPWxt) model(gwr) run(xtdpd) dgmmiv(x1 x2) * xtdpdsys spregxt y x1 x2 , nc(7) wmfile(SPWxt) model(gwr) run(xtdpdsys)

* xtmln spregxt y x1 x2 , nc(7) wmfile(SPWxt) model(gwr) run(xtmln) * xtmlh spregxt y x1 x2 , nc(7) wmfile(SPWxt) model(gwr) run(xtmlh) mhet(x1 x2) * xtfrontier spregxt y x1 x2, nc(7) wmfile(SPWxt) model(gwr) run(xtfrontier) ti * xtfrontier spregxt y x1 x2, nc(7) wmfile(SPWxt) model(gwr) run(xtfrontier) tvd * xtfrontier spregxt y x1 x2, nc(7) wmfile(SPWxt) model(gwr) run(xtfrontier) ti cost * xttobit spregxt ys x1 x2, nc(7) wmfile(SPWxt) model(gwr) run(xttobit)

选项含义为：

model(gwr) Geographically Weighted Regressions (GWR)

run( ) is used with Spatial: model(ols, sarxt, sdmxt)

主要包括如下选项模型：

7- run(xtmle) [xtreg , mle] MLE Random-Effects Panel Regression 8- run(xtam) [ NEW] Amemiya Random-Effects Panel Regression 9- run(xtbn) [ NEW] Balestra-Nerlove Random-Effects Panel Regression 10- run(xtfm) [ NEW] Fama-MacBeth Panel Regression 11- run(xthh) [ NEW] Hildreth-Houck Random Coefficients Panel Regression 12- run(xtrc) [xtrc] Swamy Random Coefficients Panel Regression 13- run(xtre) [xtreg , re] GLS Random-Effects Panel Regression 14- run(xtrem) [ NEW] Fuller-Battese GLS Random-Effects Panel Regression 15- run(xtsa) [ NEW] Swamy-Arora Random-Effects Panel Regression 16- run(xtmlem) [ NEW] Trevor Breusch MLE Random-Effects Panel Regression 17- run(xtwem) [ NEW] Within-Effects Panel Regression 18- run(xtwh) [ NEW] Wallace-Hussain Random-Effects Panel Regression

19- run(xtgls) [xtgls] Autocorrelation & Heteroskedasticity GLS Panel Regression 20- run(xtkmhomo) [xtgls] Kmenta Homoscedastic GLS AR( 1) Panel Regression * with Options: panels(iid) corr(psar1)21- run(xtkmhet1) [xtgls] Kmenta Heteroscedastic GLS - different AR( 1) in each Panel * with Options: panels(het) corr(psar1)22- run(xtkmhet2) [xtgls] Kmenta Heteroscedastic GLS - SAME/Common AR( 1) in all Panels * with Options: panels(het) corr(ar1)23- run(xtparks) [xtgls] Parks Full Heteroscedastic Cross-Section GLS AR( 1) Panel Regression * with Options: panels(corr) corr(psar1)

24- run(xtmg) [xtmg] Heterogeneous Slopes Time Series Panel Regression Requires (xtmg) module ssc install xtmg

25- run(xtpcse) [xtpcse] Corrected Standard Error Panel Regression 26- run(xtregar) [xtregar] AR( 1) Panel Regression

27- run(xtabond) [xtabond] Arellano-Bond Linear Dynamic Panel Regression 28- run(xtdhp) [ NEW] Han-Philips ( 2010) Linear Dynamic Panel Regression 29- run(xtdpd) [xtdpd] Arellano-Bond ( 1991) Linear Dynamic Panel Regression 30- run(xtdpdsys) [xtdpdsys] Arellano-Bover/Blundell-Bond ( 1995, 1998) System Linear Dynamic Panel Regression

31- run(xtfrontier)[xtfrontier] Stochastic Frontier Panel Regression

32- run(xttobit) [xttobit] Tobit Random-Effects Panel Regression

33- run(xtmln) [ NEW] MLE Random-Effects Panel Regression * Normal Model34- run(xtmlh) [ NEW] MLE Random-Effects Panel Regression * Multiplicative Heteroscedasticity Normal Model

结果为：

==============================================================================*** Binary (0/1) Weight Matrix: (49x49) : NC=7 NT=7 (Non Normalized)------------------------------------------------------------------------------==============================================================================* Spatial Panel Geographically Weighted Regression (GWR): Model(gwr) - Run(xtfe)==============================================================================* Fixed-Effects Panel Data Regression==============================================================================w1y_y = w1x_x1 w1x_x2------------------------------------------------------------------------------Sample Size = 49 | Cross Sections Number = 7Wald Test = 23.6942 | P-Value > Chi2(2) = 0.0000F-Test = 11.8471 | P-Value > F(2 , 40) = 0.0001R2 (R-Squared) = 0.3056 | Raw Moments R2 = 0.7326R2a (Adjusted R2) = 0.1668 | Raw Moments R2 Adj = 0.6792Root MSE (Sigma) = 61.4785 | Log Likelihood Function = -196.0189------------------------------------------------------------------------------- R2h= 0.3056 R2h Adj= 0.1668 F-Test = 10.12 P-Value > F(2 , 40) 0.0003 - R2r= 0.7326 R2r Adj= 0.6792 F-Test = 42.02 P-Value > F(3 , 40) 0.0000------------------------------------------------------------------------------w1y_y | Coef. Std. Err. t P>|t| [95% Conf. Interval]-------------+----------------------------------------------------------------w1x_x1 | -.2283962 .0796563 -2.87 0.007 -.3893877 -.0674047w1x_x2 | -.8133583 .2947702 -2.76 0.009 -1.409111 -.2176054_cons | 159.5689 12.69732 12.57 0.000 133.9067 185.2311------------------------------------------------------------------------------

‍

==============================================================================*** Binary (0/1) Weight Matrix: (49x49) : NC=7 NT=7 (Non Normalized)------------------------------------------------------------------------------==============================================================================* Spatial Panel Geographically Weighted Regression (GWR): Model(gwr) - Run(xtmle)==============================================================================* MLE Random-Effects Panel Data Regression==============================================================================w1y_y = w1x_x1 w1x_x2------------------------------------------------------------------------------Sample Size = 49 | Cross Sections Number = 7Wald Test = 18.9340 | P-Value > Chi2(2) = 0.0001F-Test = 9.4670 | P-Value > F(2 , 40) = 0.0004R2 (R-Squared) = 0.3055 | Raw Moments R2 = 0.7518R2a (Adjusted R2) = 0.1666 | Raw Moments R2 Adj = 0.7022Root MSE (Sigma) = 59.2296 | Log Likelihood Function = -215.7765------------------------------------------------------------------------------- R2h= 0.3055 R2h Adj= 0.1666 F-Test = 10.12 P-Value > F(2 , 40) 0.0003 - R2r= 0.7518 R2r Adj= 0.7022 F-Test = 46.45 P-Value > F(3 , 40) 0.0000------------------------------------------------------------------------------w1y_y | Coef. Std. Err. t P>|t| [95% Conf. Interval]-------------+----------------------------------------------------------------w1y_y |w1x_x1 | -.2111652 .0783551 -2.69 0.010 -.3695267 -.0528036w1x_x2 | -.7026954 .2917514 -2.41 0.021 -1.292347 -.1130438_cons | 153.0434 23.27066 6.58 0.000 106.0116 200.0751------------------------------------------------------------------------------

.

返回搜狐，查看更多