C*******I
发帖数: 339 | 1 发现最近很多IT公司都在招statistician，想问问去这些公司需要学java 或者c＋＋之
类的编程语言么？再过一年就要毕业找工作了，前两个月找实习也不是很顺利，想看看
这一年应该干点什么让自己竞争力更强一点，大牛们能给点意见么？先谢过啦～～～ |
B******5
发帖数: 4676 | 2 R and matlab for sure,
python if u have time |
L********d
发帖数: 3820 | 3 IT公司做统计又不会让你用C++，java
学好matlab，R，SAS是王道，这些都有现成的统计packages，你用java，c++不是得从
头写？
多做project，最好能做个实习比学那些东西有用多了。
【在 C*******I 的大作中提到】
: 发现最近很多IT公司都在招statistician，想问问去这些公司需要学java 或者c＋＋之
: 类的编程语言么？再过一年就要毕业找工作了，前两个月找实习也不是很顺利，想看看
: 这一年应该干点什么让自己竞争力更强一点，大牛们能给点意见么？先谢过啦～～～
|
x**********0
发帖数: 163 | 4 我觉得R 和matlab 会一个就行了，然后是SAS， SQL，
再想学excel，VBA，都挺好的
然后是python或者PHP |
B******5
发帖数: 4676 | 5 统计也用得到PHP？
【在 x**********0 的大作中提到】
: 我觉得R 和matlab 会一个就行了，然后是SAS， SQL，
: 再想学excel，VBA，都挺好的
: 然后是python或者PHP
|
x**********0
发帖数: 163 | 6 有写IT公司会要求some script programming
Python or php
一般是BI 的职位，或者data analysis
说的不对请指教哈 |
B******5
发帖数: 4676 | 7 嗯，有道理，我只是觉得PHP做网页的语言统计用不到，
不过数据库什么的估计还是需要
【在 x**********0 的大作中提到】
: 有写IT公司会要求some script programming
: Python or php
: 一般是BI 的职位，或者data analysis
: 说的不对请指教哈
|
C*******I
发帖数: 339 | 8 SAS 和Matlab倒是常用的，R以前一门课做过project 后来就再没用过，都没好意思往
简历上写，看样子还是花点时间把这三样用熟了。 |
C*******I
发帖数: 339 | 9 恩 这个听上去很靠谱 project的经验真是非常重要啊！谢谢！
【在 L********d 的大作中提到】
: IT公司做统计又不会让你用C++，java
: 学好matlab，R，SAS是王道，这些都有现成的统计packages，你用java，c++不是得从
: 头写？
: 多做project，最好能做个实习比学那些东西有用多了。
|
C*******I
发帖数: 339 | 10 是不是现在用R的比用SAS的多呀？
【在 x**********0 的大作中提到】
: 我觉得R 和matlab 会一个就行了，然后是SAS， SQL，
: 再想学excel，VBA，都挺好的
: 然后是python或者PHP
|
|
|
w********e
发帖数: 80 | |
v******i
发帖数: 1246 | 12 这个不能一概而论 depends on 这个公司做统计的人都用什么 有可能他们就是只用SAS
而已 那么对你不会有别的要求不过总的来说 sql一定要好 |
B******5
发帖数: 4676 | 13 嗯，感觉master的话学好SAS挺重要
SAS
【在 v******i 的大作中提到】
: 这个不能一概而论 depends on 这个公司做统计的人都用什么 有可能他们就是只用SAS
: 而已 那么对你不会有别的要求不过总的来说 sql一定要好
|
h******s
发帖数: 3420 | 14 the biggest challenge to get into IT is to understand indian's english
【在 C*******I 的大作中提到】
: 发现最近很多IT公司都在招statistician，想问问去这些公司需要学java 或者c＋＋之
: 类的编程语言么？再过一年就要毕业找工作了，前两个月找实习也不是很顺利，想看看
: 这一年应该干点什么让自己竞争力更强一点，大牛们能给点意见么？先谢过啦～～～
|
C*******I
发帖数: 339 | 15 恩。。。 那屁股挨着地呢？是学软件重要，还是搞学术重要啊？（想去industry的）
【在 B******5 的大作中提到】
: 嗯，感觉master的话学好SAS挺重要
:
: SAS
|
B******5
发帖数: 4676 | 16 PhD么，如果是CS的学算法重要，stat的话还是学术重要
【在 C*******I 的大作中提到】
: 恩。。。 那屁股挨着地呢？是学软件重要，还是搞学术重要啊？（想去industry的）
|
C*******I
发帖数: 339 | 17 其实是applied math，research方向是stats，现在有点困惑到底phd到industry干嘛。
。。跟master有神嘛区别？
【在 B******5 的大作中提到】
: PhD么，如果是CS的学算法重要，stat的话还是学术重要
|
B******5
发帖数: 4676 | 18 那不就是stat么，学校没有统计系？
【在 C*******I 的大作中提到】
: 其实是applied math，research方向是stats，现在有点困惑到底phd到industry干嘛。
: 。。跟master有神嘛区别？
|
l******n
发帖数: 9344 | 19 industry master和phd没有任何的区别
industry其实任何学历都没有区别，那些很老的人或许人家只有高中学历，一样做作
manager，CEO。你是不是要困惑phd到industry和高中生有什么区别？
phd是一种经历，过来的人基本都具备了独立研究问题，解决问题的能力，能够自己查
文献，自己写东西，算是高等素质教育的一部分吧
【在 C*******I 的大作中提到】
: 其实是applied math，research方向是stats，现在有点困惑到底phd到industry干嘛。
: 。。跟master有神嘛区别？
|
C*******I
发帖数: 339 | 20 恩 没有 也不发统计的phd 只有统计的master 觉得自己比起正规统计系出来的phd还是
差不少
【在 B******5 的大作中提到】
: 那不就是stat么，学校没有统计系？
|
|
|
C*******I
发帖数: 339 | 21 我明白你的意思，但是我的意思不是说phd没用，恰恰相反，我同意你说的phd是一种培
养能力的经历，我只是想知道在industry哪些工作是真正能够运用到这些能力的。不知
道说清楚了没有。
【在 l******n 的大作中提到】
: industry master和phd没有任何的区别
: industry其实任何学历都没有区别，那些很老的人或许人家只有高中学历，一样做作
: manager，CEO。你是不是要困惑phd到industry和高中生有什么区别？
: phd是一种经历，过来的人基本都具备了独立研究问题，解决问题的能力，能够自己查
: 文献，自己写东西，算是高等素质教育的一部分吧
|
g*******u
发帖数: 148 | 22 I am working at a small research company. My typical job is to bring academic ideas into the industry by implementing the methods proposed in academic papers. Because usually the authors do not share their codes, I have to replicate the papers from scratch by myself most of the time.
So far I have replicated papers in those very technical journals such as Marketing Science, Biometrika, etc. The most difficult one I've experienced which is in Management Science even requires the knowledge of reversible jump/birth-death MCMC to handle Bayesian spline regressions.
The main difficulty not only lies in the technical details, but also lies in when to stop if you smell the paper is garbage.
【在 C*******I 的大作中提到】
: 我明白你的意思，但是我的意思不是说phd没用，恰恰相反，我同意你说的phd是一种培
: 养能力的经历，我只是想知道在industry哪些工作是真正能够运用到这些能力的。不知
: 道说清楚了没有。
|
C*******I
发帖数: 339 | 23 Your job sounds interesting! May I ask what is the job title? Statistician?
Typically, what kind of companies have such job positions?
academic ideas into the industry by implementing the methods proposed in
academic papers. Because usually the authors do not share their codes, I
have to replicate the papers from scratch by myself most of the time.
Marketing Science, Biometrika, etc. The most difficult one I've experienced
which is in Management Science even requires the knowledge of reversible
jump/birth-death MCMC to handle Bayesian spline regressions.
in when to stop if you smell the paper is garbage.
【在 g*******u 的大作中提到】
: I am working at a small research company. My typical job is to bring academic ideas into the industry by implementing the methods proposed in academic papers. Because usually the authors do not share their codes, I have to replicate the papers from scratch by myself most of the time.
: So far I have replicated papers in those very technical journals such as Marketing Science, Biometrika, etc. The most difficult one I've experienced which is in Management Science even requires the knowledge of reversible jump/birth-death MCMC to handle Bayesian spline regressions.
: The main difficulty not only lies in the technical details, but also lies in when to stop if you smell the paper is garbage.
|
g*******u
发帖数: 148 | 24 the title is statistical modeler, which is equal to statistician
My industry is market research consulting. In this area, recently people are
very interested in the use of Bayesian methods to capture individual (
consumers, respondents) heterogeneity. However, I have to say there are only
few purely research-oriented positions in consulting companies. Mostly you
still need to struggle in data cleaning with SAS and SQL even if you are a
PhD.
?
experienced
【在 C*******I 的大作中提到】
: Your job sounds interesting! May I ask what is the job title? Statistician?
: Typically, what kind of companies have such job positions?
:
: academic ideas into the industry by implementing the methods proposed in
: academic papers. Because usually the authors do not share their codes, I
: have to replicate the papers from scratch by myself most of the time.
: Marketing Science, Biometrika, etc. The most difficult one I've experienced
: which is in Management Science even requires the knowledge of reversible
: jump/birth-death MCMC to handle Bayesian spline regressions.
: in when to stop if you smell the paper is garbage.
|
C*******I
发帖数: 339 | 25 This information is very helpful! Thank you so much! I think I will start
learning SQL for now. It never hurts to learn more. Right? :)
Bayesian methods seems to be popular these days. I haven only learnt the
very basic of Bayesian method, and they've always been on the side in the
statistics classes I took. What do you think is the biggest advantage of
Bayesian vs. frequentist?
are
only
you
【在 g*******u 的大作中提到】
: the title is statistical modeler, which is equal to statistician
: My industry is market research consulting. In this area, recently people are
: very interested in the use of Bayesian methods to capture individual (
: consumers, respondents) heterogeneity. However, I have to say there are only
: few purely research-oriented positions in consulting companies. Mostly you
: still need to struggle in data cleaning with SAS and SQL even if you are a
: PhD.
:
: ?
: experienced
|
g*******u
发帖数: 148 | 26 To date, Bayesian approaches can do almost everything that maximum
likelihood based methods can do, but the reverse is not true. For example,
in finance area, people had no idea of how to estimate complicated
multivariate stochastic volatility models until the introduction of Bayes
methods.
Bayesian methodology is simulation based, which means you can obtain the
whole sampling distribution as byproducts. This gives you much more
information such as higher order moments and quantiles. The main concern is
that it is computationally intensive and could take really long time due to slow
mixing. However, this is just about technical challenges. At least it is conceptually feasible for almost all problems.
In my opinion, Bayesian totally beats frequentist, the only problem is that
if the managers from old trainings are willing to accept it. To get an idea,
you can go read the article written by the Nobel Winner of this year,
Christopher Sims, which is entitled "Why Econometrics Should Always and
Everywhere Be Bayesian".
http://sims.princeton.edu/yftp/EmetSoc607/AppliedBayes.pdf
【在 C*******I 的大作中提到】
: This information is very helpful! Thank you so much! I think I will start
: learning SQL for now. It never hurts to learn more. Right? :)
: Bayesian methods seems to be popular these days. I haven only learnt the
: very basic of Bayesian method, and they've always been on the side in the
: statistics classes I took. What do you think is the biggest advantage of
: Bayesian vs. frequentist?
:
: are
: only
: you
|
l******n
发帖数: 9344 | 27 Bayesian approaches for sure have advantages over the old methods. But it
still has a lot of unsolved issues when applied in real industry, especially
in industries where certain regulations are enforced.At the same time, it
is much more expensive and time consuming to develop a Bayesian methodology
system.The cost is usually in millions and the development may take years.
I am interested in your projects, would like to know more.
Thanks,
is
to slow
conceptually feasible for almost all problems.
【在 g*******u 的大作中提到】
: To date, Bayesian approaches can do almost everything that maximum
: likelihood based methods can do, but the reverse is not true. For example,
: in finance area, people had no idea of how to estimate complicated
: multivariate stochastic volatility models until the introduction of Bayes
: methods.
: Bayesian methodology is simulation based, which means you can obtain the
: whole sampling distribution as byproducts. This gives you much more
: information such as higher order moments and quantiles. The main concern is
: that it is computationally intensive and could take really long time due to slow
: mixing. However, this is just about technical challenges. At least it is conceptually feasible for almost all problems.
|
g*******u
发帖数: 148 | 28 Yes, I concur with your comments, but as I emphasized, the real value of
Bayesian is that it makes many hopeless things before become feasible now.
Because of confidentiality, I am sorry I cannot talk too mcuh about my projects. One thing I could share is couple of weeks ago I worked on a conjunctive screening rule model which relies heavily on MCMC approaches. Following the authors' algorithm I was unable to get any sensible result and so we finally gave it up. That was the first time I understand why industry usually does not take academic research seriously.
I am just trying to provide an example that there are indeed some industrial positions requiring PhD level knowledge and training, though such positions are rare.
especially
methodology
【在 l******n 的大作中提到】
: Bayesian approaches for sure have advantages over the old methods. But it
: still has a lot of unsolved issues when applied in real industry, especially
: in industries where certain regulations are enforced.At the same time, it
: is much more expensive and time consuming to develop a Bayesian methodology
: system.The cost is usually in millions and the development may take years.
: I am interested in your projects, would like to know more.
: Thanks,
:
: is
: to slow
|
l*********s
发帖数: 5409 | 29 That is very interesting!
academic ideas into the industry by implementing the methods proposed in
academic papers. Because usually the authors do not share their codes, I
have to replicate the papers from scratch by myself most of the time.
Marketing Science, Biometrika, etc. The most difficult one I've experienced
which is in Management Science even requires the knowledge of reversible
jump/birth-death MCMC to handle Bayesian spline regressions.
in when to stop if you smell the paper is garbage.
【在 g*******u 的大作中提到】
: I am working at a small research company. My typical job is to bring academic ideas into the industry by implementing the methods proposed in academic papers. Because usually the authors do not share their codes, I have to replicate the papers from scratch by myself most of the time.
: So far I have replicated papers in those very technical journals such as Marketing Science, Biometrika, etc. The most difficult one I've experienced which is in Management Science even requires the knowledge of reversible jump/birth-death MCMC to handle Bayesian spline regressions.
: The main difficulty not only lies in the technical details, but also lies in when to stop if you smell the paper is garbage.
|
l******n
发帖数: 9344 | 30 I am not asking for details of your projects.
Bayesian approach has been advertized for long time, you can see it in many
presentations. People from academy, making some simple assumptions and doing
a demo, blow a bubble. But I have not seen any real industrial level
implementation and application. So I am curious that your company really
uses it if you replicate the paper results?
projects. One thing I could share is couple of weeks ago I worked on a
conjunctive screening rule model which relies heavily on MCMC approaches.
Following the authors' algorithm I was unable to get any sensible result and
so we finally gave it up. That was the first time I understand why industry
usually does not take academic research seriously.
industrial positions requiring PhD level knowledge and training, though such
positions are rare.
【在 g*******u 的大作中提到】
: Yes, I concur with your comments, but as I emphasized, the real value of
: Bayesian is that it makes many hopeless things before become feasible now.
: Because of confidentiality, I am sorry I cannot talk too mcuh about my projects. One thing I could share is couple of weeks ago I worked on a conjunctive screening rule model which relies heavily on MCMC approaches. Following the authors' algorithm I was unable to get any sensible result and so we finally gave it up. That was the first time I understand why industry usually does not take academic research seriously.
: I am just trying to provide an example that there are indeed some industrial positions requiring PhD level knowledge and training, though such positions are rare.
:
: especially
: methodology
|
|
|
s*********e
发帖数: 1051 | 31 实在。
many
doing
and
industry
【在 l******n 的大作中提到】
: I am not asking for details of your projects.
: Bayesian approach has been advertized for long time, you can see it in many
: presentations. People from academy, making some simple assumptions and doing
: a demo, blow a bubble. But I have not seen any real industrial level
: implementation and application. So I am curious that your company really
: uses it if you replicate the paper results?
:
: projects. One thing I could share is couple of weeks ago I worked on a
: conjunctive screening rule model which relies heavily on MCMC approaches.
: Following the authors' algorithm I was unable to get any sensible result and
|
v******i
发帖数: 1246 | 32 本来就是没区别 可是以前读过大学的人不多 你只有高中学历 也就无所谓了 几年前
统计MS还不多呢，是个统计MS都有不错的offer 现在满大街都是统计MS了 你再找工作
当然就困难了
【在 l******n 的大作中提到】
: industry master和phd没有任何的区别
: industry其实任何学历都没有区别，那些很老的人或许人家只有高中学历，一样做作
: manager，CEO。你是不是要困惑phd到industry和高中生有什么区别？
: phd是一种经历，过来的人基本都具备了独立研究问题，解决问题的能力，能够自己查
: 文献，自己写东西，算是高等素质教育的一部分吧
|
k*******a
发帖数: 772 | 33 "In my opinion, Bayesian totally beats frequentist"
我不是很清楚为什么做Bayesian的人都这么鄙视frequentist, 我记得我用的一本
Bayesian的课本的作者也是这种口气。
is
to slow
conceptually feasible for almost all problems.
【在 g*******u 的大作中提到】
: To date, Bayesian approaches can do almost everything that maximum
: likelihood based methods can do, but the reverse is not true. For example,
: in finance area, people had no idea of how to estimate complicated
: multivariate stochastic volatility models until the introduction of Bayes
: methods.
: Bayesian methodology is simulation based, which means you can obtain the
: whole sampling distribution as byproducts. This gives you much more
: information such as higher order moments and quantiles. The main concern is
: that it is computationally intensive and could take really long time due to slow
: mixing. However, this is just about technical challenges. At least it is conceptually feasible for almost all problems.
|
g*******u
发帖数: 148 | 34 My bad. Now I got you. The quick answer is, yes, people in my field are serious in the use of Bayesian . You can go check:
http://www.sawtoothsoftware.com/products/cbc/cbchb.shtml
As you can see, this is a module for the implementation of hierarchical
Bayes. The module is only 5 MB in size but asks for USD $2,000!
In the design phase we have several different methods to create surveys, while in the analytics phase the underlying method is all about random-effect logit model using hierarchical Bayes.
My projects really treat Bayes seriously. For example, I used the concept of
Albert and Chib's Bayesian residuals (1995). That paper makes extremely
good sense. In binary response models, classical residuals (Pearson,
Deviance) only help in goodness of fit test. Outlier detection is not
feasible under such traditional framework. However, with Albert and Chib's
Bayesian residuals, you can easily get the posterior distribution of the
residuals and so you can do outlier detection or things beyond that.
I think one main reason that Bayesian is not very popular in industry is
because it is more advanced both in math/stat and coding. It is not possible
to produce qualified Bayesian professionals in 2 or 3 years. Also, it is
not easy to develop a package for general use. WinBUGS has made a
significant contribution on this, but as people experience, its use is
somewhat limited. For these reasons, Bayesian is not easily tailored for industry.
Does Bayesian draw more and more attention in industry? I believe the answer is Yes. The evidence is, SAS, a conservative Stat software company, has also
realized the trends of Bayesian methods. They developed the so-called PROC
MCMC back in 2009/2010, though its evaluation is very poor.
many
doing
and
industry
【在 l******n 的大作中提到】
: I am not asking for details of your projects.
: Bayesian approach has been advertized for long time, you can see it in many
: presentations. People from academy, making some simple assumptions and doing
: a demo, blow a bubble. But I have not seen any real industrial level
: implementation and application. So I am curious that your company really
: uses it if you replicate the paper results?
:
: projects. One thing I could share is couple of weeks ago I worked on a
: conjunctive screening rule model which relies heavily on MCMC approaches.
: Following the authors' algorithm I was unable to get any sensible result and
|
C*******I
发帖数: 339 | 35 哇 趁牛人们把人气带旺了，我再问问，看到现在很多做生物统计，统计的phd都在和
master抢饭碗（比如sas programmer的工作），可是为什么我觉得phd在这些工作上并
没有明显的优势呢？已经工作的大牛们能不能给分析一下，在哪些工作职位上，phd申
请时更有优势（industry的工作）？ |
g*******u
发帖数: 148 | 36 Well, if you experienced it, you know you will never go back.
In fact, Bayesian methods have already incorporated likelihood. It strikes a balance between your prior belief and the evidence from data. Therefore, no doubt it is conceptually superior than frequentist.
There are bunches of examples. The very famous one is Kenneth Train who is
in Econ UC Berkeley. Once he experienced the power of Bayesian, he became a
Bayes advocate. He even revised his book, added more chapters, and redid all
the analysis using Bayesian methods.
【在 k*******a 的大作中提到】
: "In my opinion, Bayesian totally beats frequentist"
: 我不是很清楚为什么做Bayesian的人都这么鄙视frequentist, 我记得我用的一本
: Bayesian的课本的作者也是这种口气。
:
: is
: to slow
: conceptually feasible for almost all problems.
|
C*******I
发帖数: 339 | 37 I read the link about the talk. Honestly, I can't quite understand it. I
think it's mainly because I don't have a deep understanding of this problem
and this looks like an outline to a talk. Could you please recommend some
books for the amateurs?
a balance between your prior belief and the evidence from data. Therefore,
no doubt it is conceptually superior than frequentist.
a
all
【在 g*******u 的大作中提到】
: Well, if you experienced it, you know you will never go back.
: In fact, Bayesian methods have already incorporated likelihood. It strikes a balance between your prior belief and the evidence from data. Therefore, no doubt it is conceptually superior than frequentist.
: There are bunches of examples. The very famous one is Kenneth Train who is
: in Econ UC Berkeley. Once he experienced the power of Bayesian, he became a
: Bayes advocate. He even revised his book, added more chapters, and redid all
: the analysis using Bayesian methods.
|
g*******u
发帖数: 148 | 38 My honest suggestion is to take a class. It really takes time to enter this
area and you do need patience to digest things behind.
problem
,
【在 C*******I 的大作中提到】
: I read the link about the talk. Honestly, I can't quite understand it. I
: think it's mainly because I don't have a deep understanding of this problem
: and this looks like an outline to a talk. Could you please recommend some
: books for the amateurs?
:
: a balance between your prior belief and the evidence from data. Therefore,
: no doubt it is conceptually superior than frequentist.
: a
: all
|
c*******o
发帖数: 8869 | 39 Why Proc MCMC is evaluated poorly? Can you be more specific on that?
serious in the use of Bayesian . You can go check:
while in the analytics phase the underlying
★ 发自iPhone App: ChineseWeb - 中文网站浏览器
【在 g*******u 的大作中提到】
: My bad. Now I got you. The quick answer is, yes, people in my field are serious in the use of Bayesian . You can go check:
: http://www.sawtoothsoftware.com/products/cbc/cbchb.shtml
: As you can see, this is a module for the implementation of hierarchical
: Bayes. The module is only 5 MB in size but asks for USD $2,000!
: In the design phase we have several different methods to create surveys, while in the analytics phase the underlying method is all about random-effect logit model using hierarchical Bayes.
: My projects really treat Bayes seriously. For example, I used the concept of
: Albert and Chib's Bayesian residuals (1995). That paper makes extremely
: good sense. In binary response models, classical residuals (Pearson,
: Deviance) only help in goodness of fit test. Outlier detection is not
: feasible under such traditional framework. However, with Albert and Chib's
|
