c***y 发帖数: 460 | 1 Those rejection rates have to be taken into account for the estimation by
Sequential Probit model. Otherwise the estimates are biased. |
|
T*********I 发帖数: 10729 | 2 这里有个研究是反驳全国司法部门种族歧视的。研究的方向主要是黑人白人毒品行为的
分析。
http://www.rand.org/pubs/external_publications/EP20061001.html
Published in: Drug and Alcohol Dependence, v. 84, no. 3, Oct. 2006, p. 264-
272
A recent study of arrest data show that African Americans are 2.5 times more
likely to be arrested for marijuana possession offences than Whites, even
though general prevalence estimates show that they are no more likely to be
using. The current study investigates the purchase patterns of marijuana
users from the... 阅读全帖 |
|
F*******2 发帖数: 371 | 3 几乎所有的cdf方程都可以满足吧
logistic是general linear regression的一种,general linear regression有一个
link function,一个family function。logistic的family是binomial (也可以是
mbinomial), link function是logit。logit(p)=ln(p/1-p)
The logit model was introduced by Joseph Berkson in 1944, who coined the
term. The term was borrowed by analogy from the very similar probit model
developed by Chester Ittner Bliss in 1934.[2] G. A. Barnard in 1949 coined
the commonly used term log-odds; the log-odds of an event is the logit of
the pr... 阅读全帖 |
|
|
|
e*o 发帖数: 2134 | 6 只列举除掉本金,收益百万的,欢迎补充,自荐。
PollenAllergy: 1M
qxxxxxxxxxxg:1M
......
probit:0.5M
spanishfly:0.5M
...... |
|
R********n 发帖数: 519 | 7 Maximum likelihood就是一个求导,如果distribution不是太复杂的话,自己算算因该
很快。如果没有很解析的结果的话,那就当成一个optimization问题吧
Bayesian inference的话,那几个常用的conjugate family在很多书上都有现成公式的
,比如normal, Gamma, Beta and etc
optimize |
|
R********n 发帖数: 519 | 8 I see,Bishop的书应该是关注machine learning本身的,可能没有一些概率统计的细
节,这些在统计的书上大都都能找到 |
|
g**********y 发帖数: 423 | 9 是这篇:A Genomic Strategy to Elucidate Modules of Oncogenic Pathway
Signaling Networks
另外,我总觉得他们用binreg(svd+probit reg)来预测pathway activity有个问题,就
是他们选的gene signature based on the correlation coeff between phenotype
and the gene exp。如果你仔细看看Bild 2006 的Nature paper,可以看出每个
pathway选出的gene list数目都在变,有时候75,有时候200,300等等,然后再作svd(
效果等同PCA)。这样的feature selection 到底对不对?
就像我们做clustering,只选top highly diff expressed genes做clustering可能有
问题,也可能没事? |
|
d********m 发帖数: 3662 | 10 对SNP modeling完全不了解,是logit or probit regression + hierarchical
probabilistic model + MCMC
这个步骤吗? |
|
J**Y 发帖数: 34 | 11 My experience is that the values of latent variable from truncated normal
are easily to explode. You may want to add some truncations into your code.
Additionally, since you normalize the 1st diagnol element of covariance matrix
as 1, you can not still assume this matrix is wishart distributed. You can
check a recent paper (2000) in J. of Econometrics to see how to estimate
this kind of identified MNP model by Bayesian.
An easy way is to work on unidentified model, in which you specify priors
on |
|
k***g 发帖数: 7244 | 12 咦?一直以为 multivariate logit/probit model只有政治学家用,难道经济学家也用么
?为什么不用STATA啊,感觉要比matlab容易一些。。。 |
|
c**b 发帖数: 474 | 13 you can run the two-part model seperately by STATA, or any other software.
Actually, it contains two parts, i.e. two independent regression. Usually,
first part is a probit or logit;second part depends on your dependent
variable, which could be any regression, such as OLS, Negative Binomial
conditional on something,etc. Usually, when you run two-part model, you need
to report the results of each part. Therefore, you need to run it
seperately. |
|
c***p 发帖数: 140 | 14 假设我要估计一个discrete/continuous choice model.
第一阶段是discrete choice.
consumers face five alternatives.
I use a multinomial logit model to estimate consumers five vacation choices.
第二阶段是continuous choice. After consumers pick one choice, I would like
to know how much they spend on each choice.
我可以直接用OLS 估计吗? 假定第一个choice是去欧洲, 然后直接用OLS估计choice 1
的subsample. The dependent variable is the money spent on vacation in Europe.
我也用了Heckman selection model 估计, first stage 是 是否选择去欧洲(probit),
second stage 是 how m |
|
c***p 发帖数: 140 | 15 McFadden 和 Heckman 同获NOBEL
都是贡献在sample selection problem 上.
Heckman的两阶段模型第一阶段是probit模型.第二阶段是OLS
加入inverse mills ratio.
McFadden扩展了第一阶段是mlogit model,第二阶段也是OLS只不过加的selection term
不一样.
我主要是觉得加selection term并不难,predict the probability from mutilnomial
logit model,然后生成几个term 加进去,可是我怎么判断它是否有selection bias呢. |
|
u*********y 发帖数: 20 | 16 The dependent variable have 3 levels.
How to do the prediction? And check the Pricted yhat vs Actual y? Thanks! |
|
c*******s 发帖数: 163 | 17 Assume you are using Stata, use the "predict" command after your estimation.
You will get the predicted probabilities for each category of your
dependent variable. See Stata help file for more details. |
|
m********5 发帖数: 619 | 18 First stage
x1=z1 z2 (z1 z2 are instrumental variables)
Second stage
y=x1 x2
both estimated using OLS
第一个问题,我用SAS proc model也好,Stata ivreg也好,都会把x2也算成
explanatory variables for first stage。如果我直接用predicted value of x1来
run second stage regression,就要自己调Murphy and Topel correction. 有没有有
经验的啊
第二个问题是如果first stage estimated using OLS and second stage estimated
using probit
同样的问题,stata, ivprobit一样会把x2算成first stage里面。
第三个问题,如果distribution本身是censored, 比如volatility是>=0,这算
censored
data,需要用tobit regression |
|
t****g 发帖数: 715 | 19 First stage
x1=z1 z2 (z1 z2 are instrumental variables)
Second stage
y=x1 x2
both estimated using OLS
第一个问题,我用SAS proc model也好,Stata ivreg也好,都会把x2也算成
explanatory variables for first stage。如果我直接用predicted value of x1来
run second stage regression,就要自己调Murphy and Topel correction. 有没有有
经验的啊
ivreg y x2 (x1=z1,z2), here you go. No need to manually fix anything.
第二个问题是如果first stage estimated using OLS and second stage estimated
using probit
同样的问题,stata, ivprobit一样会把x2算成first stage里面。
Can not understan |
|
p*******t 发帖数: 501 | 20 就是binary choice logit model
顺便问,这个跟probit model的结果有本质性的区别么? |
|
X*********e 发帖数: 253 | 21 hehe, sorry,i don't know
knew very few about probit model |
|
x********4 发帖数: 405 | 22 logit assumes the error term follows log normal distribution which cannot be
negative, while probit assumes the error term follows normal distribution
that can be negative...... |
|
f*******r 发帖数: 257 | 23 Generally there are two ways of deriving the models for binary data:
1. latent variable approach: y*=x \beta + u. Here y* is a continuous
latent variable. then u can be treated as logistically or normally
distributed, which corresponds to logit or probit model. This approach is
modeling the underline variable behind the binary variable y.
2. link function approach. This approach models the binary variable y
directly: y=g(X \beta), where g() is the inverse link function. Basically
it maps |
|
|
x********u 发帖数: 411 | 25 谢谢~ 虽然读不懂,不过知道能用就好了:P
y
assumptions
anymore.
conclusion
the |
|
p********d 发帖数: 1007 | 26 you should distinguish between small sample properties and asymptotic
properties.
y
assumptions
anymore.
conclusion
the |
|
l*****o 发帖数: 61 | 27 我一组数据。dependent variable y(0,1), and independent variable X1 (
continuous variable), X2 (0为男, 1为女), and interaction term between X1 and
X2
用的是STATA 11. Model 如下:
Probit y X1 X2 X2#C.X1
Margins, dydx(*) predict(pu0)
请问(1)如何解释这种marginal effects?
(2) 如何做图X轴为X1, Y轴为predicted probability来解释X2在取值为0 和1 的区
别?
千恩万谢! |
|
l*****o 发帖数: 61 | 28 我一组数据。dependent variable y(0,1), and independent variable X1 (
continuous variable), X2 (0为男, 1为女), and interaction term between X1 and
X2
用的是STATA 11. Model 如下:
Probit y X1 X2 X2#C.X1
Margins, dydx(*) predict(pu0)
请问(1)如何解释这种marginal effects?
(2) 如何做图X轴为X1, Y轴为predicted probability来解释X2在取值为0 和1 的区
别?
千恩万谢! |
|
g*****9 发帖数: 349 | 29 刚刚查了半天statalist也没有找到答案,只能请教大家了!
我想run一个panel logistic / probit regression (binary),但是independent
variable中其中我最关注的是endogenous,所以要用IV。请问怎么能又搞IV又搞panel
logit啊?
Stata里面不知道有没有自己的或者用户编写的命令呢?
是不是这种情况只能自己run一个first stage,然后用xtlogit?但是好像这样搞
standard error就不对了。。。
真是谢谢大家了! |
|
k***g 发帖数: 7244 | 30 呵呵,那就不知道了,我见过的ordinal的dependent varible一般都是按照“strongly
agree, agree, netural, disagree, strongly disagree"这样排列的。。。你做的那个
东东以前遇到的时候,用的都是multinomial probit, 譬如以前做过一个关于alliance
type的回归分析,DV是 defence pact, netraulity or nongaggression pact和entente
,不知道能不能和你那个构成类比。。。呵呵,没有做过以人为对象的回归分析,都是以
国家或是economy,polity为对象。。。
按
Logit
都
,
里
了
do
and
protest
意
游 |
|
w****t 发帖数: 103 | 31 Given the distribution of the data, better not to run OLS. You may want to
try the following manipulations --
1) rescale your DV, say 1 - 3: 0; 4 - 5: 1; 6-7: 2 a
2) recode your DV, all values below 4 into 1, for the rest, recode into k-2
and then use ordered logit/probit
|
|
D*********Y 发帖数: 3382 | 32 【 以下文字转载自 Statistics 讨论区 】
发信人: DONTBESILLY (DONTBESILLY), 信区: Statistics
标 题: 请教credit scoring test做些什么?
发信站: BBS 未名空间站 (Thu Jun 27 14:25:02 2013, 美东)
我理解是排个序。
logit,probit,multinomial,multivariate什么的都能做scoring。cluster也算
scoring。
做credit scoring的具体做什么呢?
这是个consulting firm的position。
有什么材料可以推荐的吗?非常感谢! |
|
r***a 发帖数: 58 | 33 dependent variable Y takes either 0 or 1.
independent variable X (or Xs)
logit:
P(Y=1) = 1/ (1+ exp(-bx))
probit:
P(Y=1) = normal(bx)
the point of the model is to find beta coefficient b. |
|
r***a 发帖数: 58 | 34 这是一个值域的问题
we don't expect a probability less than zero, or greater than one, right? |
|
f*******n 发帖数: 588 | 35 呵呵,还是去统计版问一下吧。一般来说,任何一种regression
的参数估计,都有理论证明其究竟屈从于哪种分布。
Maximum Likelihood是相对于Least square而言的另一种求极值的
方法。因为这种方法通常假定随机变量屈从于z分布,所以得出的是
z statistic。但如果要严格一点去深究的话,我觉得还是需要证明。
likelihood |
|
h***i 发帖数: 3844 | 36 用LRT,不过看起来好像logit model尾巴肥一些,发生outlier的概率大一些. |
|
L*****2 发帖数: 66 | 37 hezhi 是说liklihood ratio test 去比叫那个model 好?怎么个比较? |
|
f*******r 发帖数: 257 | 38 There are two interpretations of a binary model, one is through link
function, the other one is a utility function that economists like to think
of. Think of purchasing a TV set: all you observe is a customer bought or
not. But the process behind that is a latent one: a utility function that
when reaching some value, the consumer buys. What I meant by observation at
the tail is for this utility function. If you have a large chunk of
utilities at high value or low value (such as consumers who |
|
|
o******6 发帖数: 538 | 40 ☆─────────────────────────────────────☆
iambear (iambear) 于 (Tue Feb 12 14:56:06 2008) 提到:
能举个简单例子么
☆─────────────────────────────────────☆
ctrl (ctrl) 于 (Tue Feb 12 15:05:07 2008) 提到:
http://en.wikipedia.org/wiki/Probit_model
☆─────────────────────────────────────☆
iambear (iambear) 于 (Tue Feb 12 15:15:44 2008) 提到:
it tells me nothing....
☆─────────────────────────────────────☆
yvonnefyj (yijia) 于 (Tue Feb 12 15:31:15 2008) 提到:
It is similar to logit model
In logistic regre |
|
o******6 发帖数: 538 | 41 ☆─────────────────────────────────────☆
whatsummer (不理猫@冥王星) 于 (Mon May 5 16:23:31 2008) 提到:
谢谢
☆─────────────────────────────────────☆
sotough (天马行空) 于 (Mon May 5 17:41:00 2008) 提到:
proc genmod can run poisson model also. I don't think proc logistic can run
poission.
☆─────────────────────────────────────☆
himalaya (Tea) 于 (Mon May 5 18:45:47 2008) 提到:
genmod的参数很多的 distribution可以是possion, negative binomial, normal之类
的, link可以是logit probit identity等等
http://www2.stat.unibo.it/M |
|
p*********l 发帖数: 30 | 42 The questions should be
1. what is the difference between Probit and Logit regression?
2. in what situation do we use them? |
|
l******o 发帖数: 162 | 43 is it possible that the marginal effect > 1? thanks.
10 伪币 for the first answer. |
|
s*r 发帖数: 2757 | 44 what is a marginal effect |
|
l******0 发帖数: 313 | 45 No, it should not exceed 1. |
|
|
r*****g 发帖数: 78 | 47 proc logistic data=bogus;
model y=x1 x2 /link=probit;
output out=ouput prob=p1;
run;
Y就是简单的binary, 这里output出来的p1是0到1之间的概率。有没有一个option能
output出来either 0 or 1?
我印象中好像见过,但是查sas help又好像没有。
多谢指教。 |
|
p******r 发帖数: 1279 | 48 how one unit change of indep variable can increae the log-odd of being lower
than or equal to some category, conditioned on the current value of "indep
variable"
not sure. Lou Xia continue~... |
|
|