PSM模型以及相关stata命令求教

（刚发了个一模一样的帖子但没悬赏，现在重发且悬赏回报有用的见解，如果在另一篇帖子有大神回复我了我会想尽办法把金币给大神的）
楼主在写论文分析截面数据，为了证明因果关系（且找不到合适的IV，数据本身有很大的self-selection问题），便采取了PSM模型，以及kernel matching核密度匹配。现在stata命令操作遇到了一些问题，希望有见解的朋友能分享一下解决办法，先谢过！

Outcome：formalborrowing；Treatment：cellphoneN；Covariats： capital urban hhsize Inheadagesd headage2sd headKhmerspeaking fhead headeth to15 from64 workingage headedu headpri headsec headhighcellphoneN tv car Refrigerator AirConditioner Computer Inarea roomPP floorareaPP headoccuLSOM headoccuPro headoccuTAp headoccuCle headoccuSSMS headoccuAF headoccuCT headoccuPMoA headoccuEle headoccuArm headstaEmployee headstaEmployer headstaOA headstaUP difficulty YfoodTotal YnonfoodTotal Mfoodlackstrategy
observation unit：household

遇到的问题
1.  一开始我的Treatment Variable选的是dummy"一个家庭是否拥有手机"，现在鉴于具体的手机拥有量数据可得，我便改用了ordinal variable"一个家庭拥有的手机数量"。所以我的PSM第一步stata命令也从原来的"logit treatment covariat"改成了"ologit treatment covariat, or ",从而算出propensity score
（这一步我的做法和command对吗？）
2.  匹配步骤我选取的是kernel matching，stata命令如下：
attk formalborrowing cellphoneN capital urban hhsize Inheadagesd headage2sd headKhmerspeaking fhead headeth to15 from64 workingage headedu headpri headsec headhigh tv car Refrigerator AirConditioner Computer Inarea roomPP floorareaPP headoccuLSOM headoccuPro headoccuTAp headoccuCle headoccuSSMS headoccuAF headoccuCT headoccuPMoA headoccuEle headoccuArm headstaEmployee headstaEmployer headstaOA headstaUP difficulty YfoodTotal YnonfoodTotal Mfoodlackstrategy, pscore(pscore) comsup boot detail
不是很熟悉
3.  正如我的协变量所示，非常多。主要是household occupation有11类，所以我就改成了10个dummy，看到有帖子说可以直接变成一个分类无序变量，哪位朋友可以指教一下stata的command该怎么写吗？
4.  我没用psmatch2一步到位，而是先用ologit算pscore，再用算出的pscore进行匹配，因为不确定该命令是否可以用"ologit"，可以用的话命令该怎么写呢？（之前有自己试着写了下，但结果很奇怪，如下）
>psmatch2 cellphoneN, pscore(pscore) out(formalborrowing) kernel common odds
看数据表又新生出了_pscore，且和我通过ologit算出来的pscore完全不一样，取值也超出了（0，1）这个范围，如-6.34。help看了psmatch2，不是可以直接用我已有的propensity score的吗，怎么又生出了新的？这里是团浆糊了。。。
5. PSM模型的结果变量为dummy时有什么需要注意的吗？


楼主是自学的PSM和相关stata命令，一知半解，望指教，望点拨！！贴出自己的PSM command：

>ologit cellphoneN capital urban hhsize Inheadagesd headage2sd headKhmerspeaking fhead headeth to15 from64 workingage headedu headpri headsec headhigh tv car Refrigerator AirConditioner Computer Inarea roomPP floorareaPP headoccuLSOM headoccuPro headoccuTAp headoccuCle headoccuSSMS headoccuAF headoccuCT headoccuPMoA headoccuEle headoccuArm headstaEmployee headstaEmployer headstaOA headstaUP difficulty YfoodTotal YnonfoodTotal Mfoodlackstrategy, or
>predict pscore, p
>attk formalborrowing cellphoneN capital urban hhsize Inheadagesd headage2sd headKhmerspeaking fhead headeth to15 from64 workingage headedu headpri headsec headhigh tv car Refrigerator AirConditioner Computer Inarea roomPP floorareaPP headoccuLSOM headoccuPro headoccuTAp headoccuCle headoccuSSMS headoccuAF headoccuCT headoccuPMoA headoccuEle headoccuArm headstaEmployee headstaEmployer headstaOA headstaUP difficulty YfoodTotal YnonfoodTotal Mfoodlackstrategy, pscore(pscore) comsup boot detail

​

ologit cellphoneN capital urban hhsize Inheadagesd headage2sd headKhmerspeaking fhead headeth to15 from64 workingage headedu headpri headsec headhigh tv car http://www.wxstyzy.com/ Refrigerator AirConditioner Computer Inarea roomPP floorareaPP headoccuLSOM headoccuPro headoccuTAp headoccuCle headoccuSSMS headoccuAF headoccuCT headoccuPMoA headoccuEle headoccuArm headstaEmployee headstaEmployer headstaOA headstaUP difficulty YfoodTotal YnonfoodTotal Mfoodlackstrategy, or 这段百度都翻译不出来

还有个问题：
因为要看matching得好不好，所以要比较matching前后各covariat的descriptive statistics。
原来Treatment为dummy的时候，我在一开始看covariates的descriptive statistics时简单分成cellphone＝0 和 cellphone＝1就好了。但是现在用了有序分类变量，我该如何分类看呢?感觉不是分成13种情况看的吧......

4050414 发表于 2018-12-10 11:41
ologit cellphoneN capital urban hhsize Inheadagesd headage2sd headKhmerspeaking fhead headeth to15 f ...

我用的command是：ologit treatment covariats，or
cellphoneN是我的treatment
后面的变量是协变量