nonparametric probability estimation

是 否 +2 论坛币

k人 参与回答

经管之家送您一份

应届毕业生专属福利!

求职就业群

赵安豆老师微信：zhaoandou666

经管之家联合CDA

送您一个全额奖学金名额~ !

立即领取

感谢您参与论坛问题回答

经管之家送您两个论坛币！

+2 论坛币

大家好，这是我第一次用这个论坛，要是有什么发帖格式不对请多包涵。

我用的是survey data，logistic model

我想问有没有nonparametric estimation method可以用来计算dependent variable的概率

在几篇我看过的paper上又说可以用kernel不过我不大会用啊。。然后我有好几个predictor, lpoly好像只允许一个

还有最后一个问题，在

Predictor relevance and extramarital affairs by Jeff Racine and Qi Li

中他们计算了cross validated smoothing parameters。 然后通过对比upper bound做出某些个independent variables are actually irrelevant的结论。 请问有谁知道这个是怎么做到的吗？他们用的也是kernel method, from

Cross-validation and the estimation of conditional probability densities by Peter Hall, Jeff Racine and Qi Li


谢谢！

扫码加我 拉你入群

请注明：姓名-公司-职位

以便审核进群资格，未注明则拒绝

分享0 收藏0 回帖

关键词：Probability Estimation Parametric bability Ability affairs survey method cross

Try search for the command sneop. There was a stata journal article around 2006. As for the last question, actually,  unless it happens that d someone else already knows the paper, it will be very unlikely that you can get a good answer. After all,  it is somewhat unrealistic to expect others to read a very technical paper just to post in a forum.

Also,  it would be better if you can post the two papers you mentioned, so at least if someone is interested he wouldn't need to try to find the papers himself.

I can't type Chinese here so I had to go with English...

Welcome to the forum!

Jeff racine and li qi tend to do very technical things. It is very likely that whatever they do is not available in stata. I somehow have the impression that Jeff racine uses R.
You should look at their Web pages and see if there is any replication files to download.

谢谢你的答复。当时发帖的时候没有考虑周全，不好意思。

我刚下了那个command。上面说是适用于ordered probit. 但是我的模型只是logit 也可以用吗？

erichen0112 发表于 2015-7-27 02:26
谢谢你的答复。当时发帖的时候没有考虑周全，不好意思。

我刚下了那个command。上面说是适用于ordered p ...

Sneop uses a semiparametric distribution that is very flexible. It can even be bimodal. So it does not really impose any normality assumption. I don't work in nonparametrics, but i don't quite understand why you must have logit. Wouldnt it be parametric if you impose a logistic assumption?

I didn't mean to blame you for anything at all. Just saying don't get your hopes high, because you are asking a technical question.

夏目贵志 发表于 2015-7-27 03:15
Sneop uses a semiparametric distribution that is very flexible. It can even be bimodal. So it do ...

My base model is logit, because most paper published on my topic have used linear models such as tobit and probit. However, I suspect that effects of some independent variables are nonliner. Hence, using nonparametric estimation may be one way to differentiate my research from others. So, running nonparametric estimation is more like getting a comparison group with respect to the logistic model I'm using now.

I apologise if my question seems to be stupid, as I'm not really good at econometrics or Stata.
I've checked the help page for sneop but the explanations and examples were not helpful.

I don't think nonparametrics is even a standard part of a PhD level Econometrics course in schools where no faculty member works on nonparametrics, at least not in any depth enough for research. I think the best way is to talk to your advisor and ask for help. While i do think sneop is worth trying (unless you are shooting for journals like JOE), given how little i know about this field, i don't want to be definitive.

Anyway, given modern hardware, it may not be costly to try sneop. Are you working on health or consumer/business survey?

夏目贵志 发表于 2015-7-27 09:06
Anyway, given modern hardware, it may not be costly to try sneop. Are you working on health or consu ...

Yeah I know, and thanks for the help.

The survey I'm using is General Social Survey published by NORC.

erichen0112 发表于 2015-7-27 21:55
Yeah I know, and thanks for the help.

The survey I'm using is General Social Survey published b ...

哦。是做劳动方面的啊~这个方面我实在是一无所知啊。。。