因变量为二分类变量，个体水平变量包括教育、就业等，有地区id。用两种分析方法：1）将地区作为哑变量纳入logistic回归，这样分析发现存在地区差异
2）进行二层次logistic回归，stata中命令为："melogit y x1 x2... || citynum:,or" 这里citynum就是地区id
这两种方法得到的结果如下：
方法1：
Logistic regression        Number of obs        =        1051
                                LR chi2(30)        =        231.66
                               Prob > chi2        =        0.0000
Log likelihood = -581.91416        Pseudo R2        =        0.1660

	b	ci95
survey2	1.349	[0.882,2.064]
urban	5.781***	[2.827,11.824]
turban	0.202***	[0.098,0.417]
biragegr2	1.759	[0.771,4.011]
biragegr3	1.827	[0.797,4.189]
biragegr4	1.397	[0.540,3.612]
pone	1.417*	[0.950,2.111]
ncd	0.831	[0.369,1.869]
hanmajor	0.456**	[0.249,0.837]
marital	1.762	[0.500,6.213]
edugr2	3.081**	[1.021,9.300]
edugr3	3.208*	[0.967,10.646]
edugr4	4.190**	[1.218,14.410]
job	1.058	[0.688,1.626]
incomegr2	1.408*	[0.953,2.082]
incomegr3	1.602**	[1.002,2.563]
incomegr4	1.418	[0.891,2.256]
incomegr5	1.737*	[0.967,3.121]
cleanwater	0.736	[0.449,1.206]
hygtoilet	0.863	[0.610,1.221]
htimegr	0.452***	[0.296,0.688]
nocover	0.589**	[0.374,0.926]
city2	4.229***	[2.160,8.279]
city3	0.701	[0.390,1.260]
city4	5.915***	[1.735,20.164]
city5	2.155**	[1.137,4.082]
city6	0.225***	[0.108,0.472]
city7	0.659	[0.332,1.307]
city8	3.944***	[2.065,7.532]
city9	3.060***	[1.332,7.030]
Constant	0.274	[0.056,1.343]

方法2：
Mixed-effects logistic regression        Number of obs        =        1051
Group variable:   citynum                    Number of groups=        9
Obs per group: min        =        28
avg        =        116.8
max        =        192

Integration method: mvaghermite        Integration points        =        7
Wald chi2(22)        =        84.73
Log likelihood = -599.58982        Prob > chi2        =        0.0000

	b	ci95
survey2	1.333	[0.874,2.034]
urban	5.610***	[2.784,11.305]
turban	0.218***	[0.106,0.446]
biragegr2	1.772	[0.778,4.037]
biragegr3	1.832	[0.800,4.196]
biragegr4	1.414	[0.548,3.645]
pone	1.405*	[0.945,2.089]
ncd	0.834	[0.373,1.867]
hanmajor	0.447***	[0.244,0.816]
marital	1.719	[0.490,6.030]
edugr2	3.051**	[1.013,9.188]
edugr3	3.161*	[0.956,10.457]
edugr4	4.019**	[1.173,13.771]
job	1.057	[0.689,1.622]
incomegr2	1.392*	[0.944,2.052]
incomegr3	1.568*	[0.983,2.501]
incomegr4	1.406	[0.886,2.232]
incomegr5	1.741*	[0.972,3.120]
cleanwater	0.754	[0.463,1.229]
hygtoilet	0.85	[0.602,1.201]
htimegr	0.458***	[0.302,0.694]
nocover	0.582**	[0.371,0.912]
Constant	0.441	[0.083,2.353]
var(_cons[citynum])	0.945095
Constant	2.573*	[0.976,6.781]

LR        test        vs.        logistic        regression:        chibar2(01)        =        104.93        Prob>=chibar2        =        0.0000

问题：
1）上面两种方法得到的结果比较相近，其主要的区别是不是方法1其实把地区的效应固定了，方法2把地区作为随机效应控制了？
2）从这个结果来看的话，具体该怎么判断哪种方法更加合适呢？