求助高人解答下
今天做了个关于房价的测试,数据从91年到08年的: t(时间);pri(房价);pop(人口);IR(年利率);cor(房屋建设成本);(urb)城市化程度
第一个模型是:PRI= α +β1INCt+β2POPt+β3IRt+β4CORt+β5URBt+εt
结果很糟糕,P>|t|那项显示除了cor有比较大影响,剩下的independent variable都没有什么影响,下面是结果:
. regress pri inc pop ir cor urb
Source | SS df MS Number of obs= 18
-------------+------------------------------------------------- F( 5, 12) = 5.55
Model | .09597453 5 .019194906 Prob >F = 0.0071
Residual | .041508229 12 .003459019 R-squared = 0.6981
-------------+----------------------------------------------- Adj R-squared = 0.5723
Total | .137482759 17 .008087221 Root MSE = .05881
------------------------------------------------------------------------------
pri | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
inc | .07411 .44087 0.17 0.869 -.8864633 1.034683
pop | 4.480694 35.3166 0.13 0.901 -72.46756 81.42894
ir | -.0036233 .0082428 -0.44 0.668 -.0215827 .0143362
cor | .7221647 .2370482 3.05 0.010 .2056811 1.238648
urb | -.1003941 1.372459 -0.07 0.943 -3.090726 2.889938
_cons | .0469459 .786432 0.06 0.953 -1.666542 1.760434
------------------------------------------------------------------------------
然后用了第二个模型:lnPRI= α +lnβ1INCt+lnβ2POPt+lnβ3IRt+lnβ4CORt+lnβ5URBt+εt
然后就结果就好多了
regress logpri loginc logpop logir logcor logurb
Source | SS df MS Number of obs= 8
-------------+------------------------------ F( 5, 2) = 19.08
Model | 3.22763639 5 .645527277 Prob >F = 0.0506
Residual | .067671343 2 .033835671 R-squared = 0.9795
-------------+------------------------------ Adj R-squared = 0.9281
Total | 3.29530773 7 .470758247 Root MSE = .18394
------------------------------------------------------------------------------
logpri | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+------------------------------------- ---------------------------
loginc | 1.165712 .2963565 3.93 0.059 -.1094067 2.440831
logpop | 10.09308 2.080364 4.85 0.040 1.141994 19.04416
logir | -.2454173 .2198075 -1.12 0.380 -1.191173 .700338
logcor | .2216548 .104405 2.12 0.168 -.2275636 .6708731
logurb | 17.14887 3.925733 4.37 0.049 .2578032 34.03993
_cons | 67.6519 14.33035 4.72 0.042 5.993377 129.3104
------------------------------------------------------------------------------
我觉的我用的是OLS模型,不存在异方差(散点)问题啊,但是log确实修正了P>|t| (significant)这个值
用第一个模型的时候同时还存在serial correlation(序列相关?不知道中文是什么)的问题,用了第二个log模型之后又修正了
我懒,直接用DW测试(durbina)结果如下:
模型一:
. durbina
Durbin's alternative test forautocorrelation
---------------------------------------------------------------------------
lags(p) | chi2 df Prob > chi2
-------------+-------------------------------------------------------------
1 | 1.608 1 0.2047
---------------------------------------------------------------------------
H0: no serial correlation
存在serial correlation 因为 Prob>chi2 这项大于0.1
模行二:
Durbin's alternative test forautocorrelation
---------------------------------------------------------------------------
lags(p) | chi2 df Prob > chi2
-------------+-------------------------------------------------------------
1 | 3.374 1 0.0662
---------------------------------------------------------------------------
H0: no serialcorrelation
这个模型就小于0.1了,问题解决了。为什么呢????