Z***U
发帖数: 216 | 1 个人觉得如果
1） Correlation matrix 的 off diagonal elements较大，也就是具有一定的 correl
ation
2） Factor 的数量选择相同的话， loading 给出的解释应该是大致差不多的，虽然具
体的数值可能会有差别。
是的，负值是在 prinit 或者 ml 时才可能得到的，因为iteration之后再用来
decompose的communality 矩阵可能不再是 positive definite。
因此，prin 更多的解释了standardized responses 的 variance，
而 prinit 和 ml 更多的考虑到 correlation，所以 loading 会有不同，但是如果数据
确实具有较强的correlation，比如 MSA 较大，那么所得的所有factor 的解释不会有
很大差别。 不过个别的 factor 则会有很大差别。 |
|
z****e
发帖数: 2024 | 2 【 以下文字转载自 Programming 讨论区 】
发信人: SLWY (松林物语), 信区: Programming
标 题: 一个比较有趣的面试问题
发信站: BBS 未名空间站 (Tue Dec 8 01:51:17 2009, 美东)
觉得比较有趣。
鸭子喜欢在冬天晒太阳。 鸭子可能一天呆在一个草地的几个地方晒太阳。比如一开始在
坐标轴是 (1,2)的地方晒， 后来移动到 (10,10), 再后来移动到 (7,6). 研究发现，
鸭子天与天之间选的地方有correlation.比如，鸭子今天选的是
(1,2), (10,10), (7,6)
那么明天选的地方很可能离这些很近:
(1.1,2.3),(11, 10.5), (7.7,6)
发现有些鸭子correlation大些，有的小些。现在给你几天的数据，让你develop一个函
数，或者model来量化这些天与天之间的correlation（一天当中的地点不用考虑correl
ation,也就是(1,2)与(10,10),(7,6)之间的cor |
|
f****7
发帖数: 398 | 3 请教大家一个关于stepwise programming的问题,多谢大家宝贵的时间：
有一个dataset含有101个变量，现在希望依次找出前100个变量中哪些和最后一个变量
的有相关关系。解决的步骤是
1.找出correlation最高的var1，将var1与前100个变量中剩下99个variable分别相加，
得到一组新的variables
2.从新的99个变量中找出correlation最高的那个变量var2, 此var2实际是原dataset中
的var1+var2，并将var1,var2与原100个变量中剩下98个变量相加，得到一组新的
variables
3.从新的98个变量中找出correlation最高的变量var3=（var1+var2+var3）
4.重复以上步骤，直到每次运算中最高的correlation与前一次相比不再增加。
我对此类编程没有经验，现在还徘徊在到底是用sas还是r来处理的阶段，希望大家多多
指教，不甚感激。 |
|
p********a
发帖数: 5352 | 4 ☆─────────────────────────────────────☆
fff007 (felix) 于 (Mon Mar 8 15:09:01 2010, 美东) 提到:
请教大家一个关于stepwise programming的问题,多谢大家宝贵的时间：
有一个dataset含有101个变量，现在希望依次找出前100个变量中哪些和最后一个变量
的有相关关系。解决的步骤是
1.找出correlation最高的var1，将var1与前100个变量中剩下99个variable分别相加，
得到一组新的variables
2.从新的99个变量中找出correlation最高的那个变量var2, 此var2实际是原dataset中
的var1+var2，并将var1,var2与原100个变量中剩下98个变量相加，得到一组新的
variables
3.从新的98个变量中找出correlation最高的变量var3=（var1+var2+var3）
4.重复以上步骤，直到每次运算中最高的correlation与前一次相比不再增加。
我对此类编程没有经验，现在还徘徊在到底是用sa |
|
G**7
发帖数: 391 | 5 The predictor variables are ordinal.
Why not read some literature to determine first?
Next, get the correlation of all variables and if any are highly correlated
or not correlated. Double check with literature for those with 0
correlation. |
|
k****i
发帖数: 347 | 6 为了5个包子，上了
-----------两组间在 Month 1的mobility结果上有没有区别?
用t-test (continuous outcome) 或 chi-square (categorical outcome) 就可以了
proc ttest;
proc freq;
-----------How about across both Month 1 and Month 12?
考虑到是repeated measure，用linear mixed model（对continuous outcome）
或者generalized linear mixed model（对dichotomous或categorical outcome）
proc mixed data=;
class month id;
model mobility = treatment / solution;
repeated month / subject=id type=un;
run;
------------两组间在 Month 1的memory结果上有没有区别? How about across... 阅读全帖 |
|
|
u****m
发帖数: 118 | 8 统计分描述性统计,例如mean,variance, SD之类都是对样本的description,没有对应的
significant test.
另一类就是建立在这些样本参数之上的推断性统计,需要假设检验,significant test,
结论都是对总体下的.
correlation应该是属于第二者,因此我认为对一个人的数据做correlation是没有意义
的,即便是repeated measure.
不过我导师要我在写paper的时候对一个人数据做correlation,事先说明我的专业不是
统计,也不是在商学院,而是工程学院,需要用到统计知识做研究,但是又不如其他学院精
通统计.
所以上统计版来问问就一个人数据做correlation到底对不对. |
|
y******i
发帖数: 558 | 9 Correlation is only "partially" meaningful when you have a large sample set.
How can a single gene mutation "correlates" to another historical
development?
The term "correlates" here isn't referring to anything statistical. |
|
k**********4
发帖数: 16092 | 10 How to Interpret a Correlation Coefficient r
By Deborah J. Rumsey from Statistics For SB, 2nd Edition
In statistics, the correlation coefficient r measures the strength and
direction of a linear relationship between two variables on a scatterplot.
The value of r is always between +1 and –1. To interpret its value, see
which of the following values your correlation r is closest to:
Exactly –1. A perfect downhill (negative) linear relationship
–0.70. A strong downhill (negative) linear rel... 阅读全帖 |
|
c***z
发帖数: 6348 | 11 Do you know calculus?
Say you have a function y = f(x), where f has some parameters, and follow
some basic requirement (e.g. continuous, but this is not necessary, because
saw shaped functions work better).
Basically it is using the chain rule in calculus to move the parameters
around for minimal prediction error, which is the difference between the
predicted y and real y.
When error is minimized, we think we found a correlation between the inputs
and outputs, and try to apply that correlation t... 阅读全帖 |
|
d**********i
发帖数: 524 | 12 We did find a significant association between political conservatism and
rs10952668 (Table 5). This SNP lies in LOC642355, a pseudogene on chromosome
7. Not surprisingly, the SNP also showed an association with the highly
correlated trait of Democrat versus Republican (b=0.260, P<
.02).
An interesting finding was that the SNP associated with political
conservatism, rs10952668, also showed marginal evidence for association with
the personality traits openness (b=0.142, P<.06) and
ag... 阅读全帖 |
|
h*******g
发帖数: 2201 | 13 Logically that is flawed. First of all, there must be many other things
correlated with the economy development of states, and this correlation does
not justify this behavior is good or bad. You even see more waste produced
from more developed countries.
Secondly this correlation, even if true, only proves the larger attraction
of gays to highly developed states, and still it does not give us an answer
to the question: What causes gay? |
|
s**********e
发帖数: 2888 | 14 影帝Colin Firth，国王的演讲男主角，去年在Current Biology上面发表一篇文章，讨
论左派右派的人的大脑结构不一样。
http://www.ncbi.nlm.nih.gov/pubmed?term=colin%20firth
Political orientations are correlated with brain structure in young adults.
Kanai R, Feilden T, Firth C, Rees G.
Source
University College London Institute of Cognitive Neuroscience, 17 Queen
Square, London WC1N 3AR, UK. r*****[email protected]
Abstract
Substantial differences exist in the cognitive styles of liberals and
conservatives on psychological measures. Variability in pol... 阅读全帖 |
|
f**********n
发帖数: 29853 | 15 不可能有官方数据，因为政府不记录宗教信仰。
这是维基页面
Religion
A few studies have found a negative correlation between religiosity and
criminality. A 2001 meta-analysis found, "religious beliefs and behaviors
exert a moderate deterrent effect on individuals' criminal behavior".[5][6]
An individual with high religious saliency (i.e. expressing the high
importance of religion in their life) is less likely to be associated with
criminal activities; similarly, an individual who regularly attends
religious services or is hig... 阅读全帖 |
|
|
b******e
发帖数: 82 | 17 Can someone explain the rationale of the correlation between the likelihood
of a proposal rejection and the change of status date shortly after the
panel review? I heard about this correlation, but I just do not understand
the rationale of this correlation. Someone said that the change of the
status date means that the PM has uploaded the review comments to the
fastlane server. I wonder that the review comments for the proposals
receiving HC or C should also be uploaded to the fastlane server.
T... 阅读全帖 |
|
G**7
发帖数: 391 | 18 Hi, friends:
You are all very smart.
I just have a quick question. I have two events--(1) 1st year, 2nd year,
3rd year and 4th year registration records of some high school students and
(2) students ever participating in some science club during their high
school years.
Can we correlate (1) with (2) or vice versa?
The interest is the impact of science club on high school graduation and
registration. However, we don't know when the students participated in the
science club at all.
We ... 阅读全帖 |
|
t****r
发帖数: 702 | 19 it's called Multicolinearity. It is probabily because some columns of the de
sign matrix (age, sex, age*sex) have strong correlations (linearly dependent
),
which inflates standard errors of the coefficient estimator, leading to
insiginificant p-values. In otherwords, age*sex may have contained similar
information as those in age and sex, hence once you include age*sex, then
age and
sex become redundant. You can check pairwise correlations between age, sex
and age*sex to see if there is strong c... 阅读全帖 |
|
t****r
发帖数: 702 | 20 it's called Multicolinearity. It is probabily because some columns of the de
sign matrix (age, sex, age*sex) have strong correlations (linearly dependent
),
which inflates standard errors of the coefficient estimator, leading to
insiginificant p-values. In otherwords, age*sex may have contained similar
information as those in age and sex, hence once you include age*sex, then
age and
sex become redundant. You can check pairwise correlations between age, sex
and age*sex to see if there is strong c... 阅读全帖 |
|
a***r
发帖数: 146 | 21 2012 markets: Expect ups and downs
BY Jurrien Timmer, Co-Manager of Fidelity® Global Strategies Fund,
Fidelity Viewpoints — 12/21/11
Consider U.S. stocks, high-yield corporate, and floating-rate high-income
bonds.
It’s been quite a year, one of violent mood swings but little overall
direction. We seem to be in a time warp where everything happens faster and
faster. Everything seems to be correlated. There are very few places to hide
, and even those places don’t feel like good options anymor... 阅读全帖 |
|
s*****e
发帖数: 21415 | 22 股市里的人一般都健忘，特别inpressionable。类似小鸭子，基本上生下来看到
谁就认谁当妈。
牛市火了几年就觉得股票市场或者房市一直牛；熊市熊了几年就觉得经济很快
要崩溃，世界很快要末日；黄金火了几年就有人叫嚣要回复金本位；如此种种
奇谈怪论，不一而足。
现而今有index至上论。平心而论，index是一种不错的投资方式。不过do better
也不像很多人说的那样near impossible。
前几年世界经济动荡的很，从美国金融危机开始，到后来的欧债危机一会儿一抽
风，基本上整个市场correlation很大，要涨一起涨，要跌一起跌。这种状况持续
了几年，所以看起来index投资是最靠谱的，单个股票的分析和选股价值不大。这
也是现在index至上论的根源。
不过，如果说我们对市场足够了解的话，things always change，现在correlation
大，不代表将来correlation大。其实index投资惨的时候也不是没有过，2000年
internet泡沫破裂的时候，talk to the indexers! 包括08年危机的时候，你要是
手握index满仓... 阅读全帖 |
|
M********r
发帖数: 278 | 23 You are right that correlations among assets vary over time. If you think
now is bad, consider the 70's when the correlation btw US stock and US bonds
was around 0.6-0.7 instead of a negative number. This simply reflects the
randomness in the returns of these assets. In the long term, however, bonds
are still effective diversifiers against stocks. Also, we should consider "
magnitude", that is when assets become correlated than usual, the magnitude
of price movements differs. Almost always, bond... 阅读全帖 |
|
D***r
发帖数: 7511 | 24 增长rp
半路出家CS硕士，学的是machine learning方向的粗浅技能，主要是computer vision
那个公司是搞NLP还有搜索什么的，方向不是很对口但也不能说无关
他上来就问怎么对一些文本的entries进行分类
这个我还是大概知道，就说可以用key words，有naive bayes之类的方法
然后他就追问training data大概需要多少词
这纯粹是经验问题，弄过的才知道。我当然不知道，就说反正需要做feature
selection
然后他问我英语一共有多少词，我说常用的几万吧。
他就追问怎么进一步减少
我知道要做feature selection，但是不知道这种数据一般是怎么做
就按照图像处理用到的思路瞎说了一气，比如在feature之间找correlation
然后de-correlate数据
反正虽然不着调，但从原理上肯定是没错
（后来反应过来词之间的correlation有很好的定义，比如同义词，不用数学方法来找）
然后他问怎么把一些词根相同的词合并
我说了morphological analysis和string comparison
都不是他... 阅读全帖 |
|
a*****g
发帖数: 19398 | 25 前一段，我把“围棋数学”的构思和进展情况介绍给了一位教授。
这位教授已经做了二十多年数学和工程的教育工作。
教授今天给回信——最后一句太鼓舞人心了——翻译成中文就是——
“如果你解决了2个难题，围棋数学将会席卷美国的每一个学校……”。
【 I do think that you are correct in noticing a correlation between math an
d go. Certainly, basic arithmetic... adding, subtracting, multiplying and th
eir geometric analogues are all there at the lowest levels of the game. But
also, estimating is there. And that is more abstract. Then the bigger, and
more interesting, question is whether or not the planning that comes up in
G... 阅读全帖 |
|
a*****g
发帖数: 19398 | 26 前一段，我把“围棋数学”的构思和进展情况介绍给了一位教授。
这位教授已经做了二十多年数学和工程的教育工作。
教授今天给回信——最后一句太鼓舞人心了——翻译成中文就是——
“如果你解决了2个难题，围棋数学将会席卷美国的每一个学校……”。
【 I do think that you are correct in noticing a correlation between math an
d go. Certainly, basic arithmetic... adding, subtracting, multiplying and th
eir geometric analogues are all there at the lowest levels of the game. But
also, estimating is there. And that is more abstract. Then the bigger, and
more interesting, question is whether or not the planning that comes up in
G... 阅读全帖 |
|
l******a
发帖数: 16364 | 27 ☆─────────────────────────────────────☆
northbeach ((想认识一个小时候放过羊的朋友)) 于 (Tue Mar 24 16:27:55 2015, 美东) 提到:
“在美国, 所谓“affirmative action” 又叫“平权法案”，指的是“平等人权在就业
领域的体现”，即 “equal opportunity employment”。具体而言，“平权法案”要
求美国联邦的各级 contractors and subcontractors必须遵照执行。 其目的是杜绝各
类就业歧视，包括因肤色，信仰，性别，国别等引起的歧视。”
上面是关于ＡＡ的解释，来源来自旁边的楼。从字面上，ＡＡ绝对是个好东西，让各个
种族都能被平等地对待。 同时，ＡＡ在美国历史上正面作用不可抹杀。在美国，曾经
有色人种不能和白人坐一样的置位，有色人种不能上白人的学校。种族歧视的受害者绝
不是只是黑人，当时杨振宁贵为诺贝尔奖得主，也不能在白人区里面买房子。正是ＡＡ
，然后学校，雇主，不能名目张胆地以种族的理由来拒绝某个人。
但是任何事情都有里面性。因为Ａ... 阅读全帖 |
|
l******a
发帖数: 16364 | 28 ☆─────────────────────────────────────☆
northbeach ((想认识一个小时候放过羊的朋友)) 于 (Tue Mar 24 16:27:55 2015, 美东) 提到:
“在美国, 所谓“affirmative action” 又叫“平权法案”，指的是“平等人权在就业
领域的体现”，即 “equal opportunity employment”。具体而言，“平权法案”要
求美国联邦的各级 contractors and subcontractors必须遵照执行。 其目的是杜绝各
类就业歧视，包括因肤色，信仰，性别，国别等引起的歧视。”
上面是关于ＡＡ的解释，来源来自旁边的楼。从字面上，ＡＡ绝对是个好东西，让各个
种族都能被平等地对待。 同时，ＡＡ在美国历史上正面作用不可抹杀。在美国，曾经
有色人种不能和白人坐一样的置位，有色人种不能上白人的学校。种族歧视的受害者绝
不是只是黑人，当时杨振宁贵为诺贝尔奖得主，也不能在白人区里面买房子。正是ＡＡ
，然后学校，雇主，不能名目张胆地以种族的理由来拒绝某个人。
但是任何事情都有里面性。因为Ａ... 阅读全帖 |
|
G**7
发帖数: 391 | 29 Hi, friends:
You are all very smart.
I just have a quick question. I have two events--(1) 1st year, 2nd year,
3rd year and 4th year registration records of some high school students and
(2) students ever participating in some science club during their high
school years.
Can we correlate (1) with (2) or vice versa?
The interest is the impact of science club on high school graduation and
registration. However, we don't know when the students participated in the
science club at all.
We ... 阅读全帖 |
|
|
g****u
发帖数: 695 | 31 All right, part 2, Technical Analysis.
I threw out a little hint at the end of part 1 by saying TA doesn't
work, and it got some defensive words back already. It will be
interesting to see what happens after I finish this part, because
if you accept what I am going to say, you probably will have to start
over in your trading adventure (which I think you should).
It is worth a few words to talk about what TA is trying to solve
first. Market price, no matter what the underlying product is, all
com... 阅读全帖 |
|
P**H
发帖数: 2272 | 32 用SPX rally中moving correlation权重tickets的来分析是不是也会得到一样的结果？
但是每个的moving correlation可能有个周期变化。我注意到很多出现周期性的正相关
，0相关和负相关变化，感觉如果用100日或者200日长均线找moving correlation的权
重，然后计算每个的小周期，然后用10日或20日的数据加权累加可能更靠近市场的方向。 |
|
m********0
发帖数: 2717 | 33 I was over-reacting, sorry.
I agree with you consistency is the key. And I always believe in consistency
makes ATM.
Market evolves, a lot strategies works well does not work any more.
Meanwhile, there are more strategies been created, more systematic framework
, more computing power to search patterns, even patterns like short X every
trading day around XX:XX(maybe combined with market up/down for that opening
). I have a few candidates in my mind.
And there are different *market inefficiency* i... 阅读全帖 |
|
p******e
发帖数: 17163 | 34 闲来无事，发个贴谈一下黄金。 事先声明，以下用来估黄金价值的方法不适宜用来炒
黄金，也不是本人用来炒黄金的基础，但对于长线投资或许有点参考价值。
1）找到1970年以来每月美联储的mb,m2数据，以及美国的cpi数据。 再找到1970年以来
的每月黄金价格。
2）取上面数据12月移动平均值。
黄金对mb correlation 为85%， 对m2为79%，对cpi为73%
（其实黄金对m2的correlation与黄金对mb的correlation在金融危机前是一样的，因为
在require reserve ratio 为十的情况下， m2:mb 的比例一直固定在10：1， 而金融
危机以来，mb大幅增长，但m2 没跟上来，（还好没跟上来，要跟上来就
hyperinflation咯）
3）分别regress黄金对mb, m2 以及cpi。
假设明年没有qe, mb只会增长一点， 明年移动平均值大约在2.7万亿左右。
假设m2以过去3个月的速度增长， 那明年移动平均值不会超过10万亿。
假设cpi增长3%，那明年cpi移动平均值不会超过235。
根据以上假设， 明年黄金平均价格
以mb估... 阅读全帖 |
|
c*******y
发帖数: 1630 | 35 最后再说一下那个quant fund的research paper，有一篇标题叫
Unified Risk Theory - Correlation, Vol, M3, and Pineapples
我觉得挺有深度，
基本上讲了不要静态地去理解market，现在由于大量的高频交易和ETF，使得各种产品
，甚至不同领域的关联性大大增加。现在一个stock market的消息引起的指数暴跌会可
能牵连原本比较强劲的大宗商品板块。文中有很多通俗易懂的统计数据，我就不列了。
一句话，correlation的增加，会导致volativity的volatilty增加。而当QE1,QE2出来
时候，我们会看见VIX稳步下降，直到出现一些例如地震，地缘真挚等等重大利空。另
外作者不认为correlation的增加的本质原因是ETF和HFT。他觉得是货币供给引起的全
球性的市场结构变化。
文中还举了一个伯南克和夏威夷水果市场的小例子。很有趣。
不多说了，VIX产品越来越多，现在CBOE已经推出了VIX的VIX了。如果按照文中的意思
是，可以BAFD。VIX的TERM STRUCTURE跟以前也很不一样了... 阅读全帖 |
|
T**7
发帖数: 264 | 36 杂志介绍
http://www.rsc.org/publishing/journals/MB/about.asp
Impact factor: 3.859
劳驾站内信发给我：
你的名字，你的email，你的专业特长关键词（3－5个即可），你的发表文章。
为了你省事，你可以copy paste你的resume也可以
希望不同专业都来试试，“分子生物系统”杂志算是交叉学科杂志。
谢谢！
稿件摘要：
TITLE: Noninvasive measurement of respiratory activity of mouse embryoid
bodies and its correlation with undifferentiation/differentiation markers
ABSTRACT: Mouse embryoid bodies (mEBs) were evaluated in detail on the basis
of respiratory activity and high-throughput quantitative reverse
transcriptio... 阅读全帖 |
|
l*******c
发帖数: 523 | 37 据烙印论坛说，有人PD是09-11-2008的已经收到NVC的 Fee notice.
Reference: http://www.qesehmk.org/forums/showthread.php?148-EB2-Predictions-(Rather-Calculations)-2012/page112
以下是烙印对一月份VB的分析，感觉比较合理：
Following is the current trend based on the dashboard, Trackitt and other
news and information.
GC Issuance date based on 25K will be 15-DEC-2007 - 01-JAN-2008 by the end
of FY 2012.
Next VB (Jan 2012) in all probability will be the final intake if HR-3012
does not go through.
Possibility 50% - 2 months of additional int... 阅读全帖 |
|
c**i
发帖数: 6973 | 38 I apologize. Regarding the last paragraph in my yesterday's posting. The
critical p should be 0.05 rather than 0.5. *That is why yesterday I
mentioned 1 out of 20, which is 0.05.
The p is a number between 0 and 1. The larger the p, the more likely the two
things are correlated.
* * *
My comment: In statistics, the probability p means the chance that two
things are correlated. When p is equal to 0.5, that means one out of twenty
times (0.5), the two things appear correlated ALTHOUGH they are not |
|
a*****g
发帖数: 19398 | 39 前一段，我把“围棋数学”的构思和进展情况介绍给了一位教授。
这位教授已经做了二十多年数学和工程的教育工作。
教授今天给回信——最后一句太鼓舞人心了——翻译成中文就是——
“如果你解决了2个难题，围棋数学将会席卷美国的每一个学校……”。
【 I do think that you are correct in noticing a correlation between math an
d go. Certainly, basic arithmetic... adding, subtracting, multiplying and th
eir geometric analogues are all there at the lowest levels of the game. But
also, estimating is there. And that is more abstract. Then the bigger, and
more interesting, question is whether or not the planning that comes up in
G... 阅读全帖 |
|
t**********p
发帖数: 2672 | 40
Again, why you use statistics to explain and support your conclusion? A
correlation is a necessary but not a sufficient condition alone to make
causal inferences with reasonable confidence to make conclusion.
This is more complicate than you think, summarizing with one sentence
explain very well your strong biased against Christian, how about share with
us your negative experience or personal struggle with Christian or God?
"Again, there's quite a bit of variation from state to state, but overa... 阅读全帖 |
|
T*********I
发帖数: 10729 | 41 NJ一个州以郡为单位的数据毕竟是少数。如果你拿全美国以州为单位，或者全世界以国
为单位，那么结果就要有说服力的多。
这上面两种我做过，所以给你的推荐是：
自杀率在美国和大锅饭有关（gini，负相关）；在全世界和IQ有正相关（越聪明，越自
杀）。
死亡率（因为暴力引起的，包括杀人、强奸、抢劫和攻击导致的死亡）无论在美国，或
者在全世界都和IQ有负相关。
上述两个因素都是最大的。
我注意到你的帖子里面用的仅仅是Murder。而murder仅仅是暴力犯罪死亡的一部分（如
jerseyv所说，murder： the unlawful premeditated killing of one human being
by another.)FBI的数据要更全些，包括了其他的犯罪导致死亡的数目。
咱们就以你说的Murder来看，IQ仍然是最相关的。而枪仍然是最不相关的。而且还是负
相关--震慑
IQ correlation with
Murders Rate -0.630674166
Gun Murder Rate -0.554348063
... 阅读全帖 |
|
t*******r
发帖数: 22634 | 42 那些个研究还只是在医学统计学阶段，统计上看看有没有 correlation 而已。
但即使有 correlation，很有可能也没有 causation。正如你说的，由于
其他非相关因素的影响。
但另一方面，没有证据表明 correlation，也不能排除有 causation。
有可能样本不够，有可能被其他非相关因素（类比于噪音）掩盖。
不过俺个人看法 IVF 总是（1）涉及到使用激素类药物（2）绕过了受精
过程对精子/卵子的自然选择。要断言说肯定不会有任何影响，可能也难。
目前的医学其实还很原始。要俺老农说，其实没那么复杂。如果能克服
困难自然受孕，当然 IVF 想都不要去想，麻烦点就麻烦点。如果不行，
IVF 也不用怕，那么多人都搞出健康宝宝了，所谓 birth defect 高或
低，其实是对理论研究和技术进步的意义更大些，对个案说白了主要还
是看 rp ………… |
|
w*********m
发帖数: 4740 | 43 工资高or high median household income is an observation, it has some
correlation with food quality, but not the root cause of food quality.
the root cause is 供求关系. and high median household income has some
correlation with 供求关系 when other factors are the same or similar.
so given other factors are the same or similar, high median household income
has negative correlation with food quality with the same price.
food quality does not refer to the restaurant environment. |
|
l*****a
发帖数: 3630 | 44 季后赛开始了，好多著名投手一上来就被打爆，Severino, Sale, Kluber, Kershaw,
Greinke. 有感而发，顺便谈谈棒球中的运气。
棒球里一个重要的概念就是安打，那么什么是安打? 简单说就是打出的球不能在落地之
前被野手接到，同时打手达到一垒前球没传进一垒手的手套。看球时经常会看到这样的
例子，一个打手打出一个很强劲的平飞球(line drive)， 但是由于位置不好，直接被
接杀。有时候打出一个软弱无力的滚地球，却可能形成内野安打。所以，安打其实是一
个很复杂的概念，同时，野手的站位，野手的守备能力等都会影响安打的几率。那么如
何来描述安打几率的这个”运气”值呢，于是就有了BABIP. BABIP=(H - HR)/(AB - HR
- K + SF), 简单说就是去掉保送，三镇，全垒打后，球打出球场内形成安打的几率。
大数据下，BABIP这个值大约为0.3, 也就是说统计意义上10个球平均有3个会变成安打
。如果一个投手BABIP过低(比如今年上半赛季的Erwin Santana，大约只有0.2)，一般
可以认为该投手的“运气”比较好，统计意义上此人的成... 阅读全帖 |
|
a*****g
发帖数: 19398 | 45 前一段，我把“围棋数学”的构思和进展情况介绍给了一位教授。
这位教授已经做了二十多年数学和工程的教育工作。
教授今天给回信——最后一句太鼓舞人心了——翻译成中文就是——
“如果你解决了2个难题，围棋数学将会席卷美国的每一个学校……”。
【 I do think that you are correct in noticing a correlation between math an
d go. Certainly, basic arithmetic... adding, subtracting, multiplying and th
eir geometric analogues are all there at the lowest levels of the game. But
also, estimating is there. And that is more abstract. Then the bigger, and
more interesting, question is whether or not the planning that comes up in
G... 阅读全帖 |
|
G***a
发帖数: 27294 | 46 有很大的ｃｏｎｆｏｕｎｄ啊！！
37岁超重的白人（２４．５－－２９），很有可能年轻时候ＢＭＩ正常。
那么假设其中有两拨人，一拨人是Ａ，一拨人是Ｂ。假设各有２０个人。
他们年轻时候，都正常ＢＭＩ，假如都活着，中年以后都超重。
假定Ａ组本身就有很大早死的ｒｉｓｋ，比如遗传病，或者爱冒险，吸毒啥的
假定Ｂ组人，早死ｒｉｓｋ低，没遗传病，生活老实健康。这种情况下，
按他们方法，从２５岁ｆｏｌｌｏｗ１２年，Ｂ组人全活下来了。Ａ组死了假设５个人
那么Ｂ组的人的平均体重按这２０个人，３７岁时候体重算
Ａ组呢？15个人按37岁算，剩下的人的体重，肯定算20多岁/30出头死的时候得体重吧
？
那可不就得出来，超重的人活得长嘛！～ （白人２５岁到３７岁，体重增长很快的）
简单的总结，体重和年龄大大correlate, 年龄又和生死大大correlate
如果一个variable，和independent variable和dependent variable都correlate
就是confound ！！！！ ----Statistic 101 ！！！
ages
mass |
|
P****D
发帖数: 11146 | 47 correlation（动词为correlate）描述dependence。correlation是所有statistical
relationship（动词为relate）中的一种。 |
|
K***a
发帖数: 3614 | 48 【 以下文字转载自 Statistics 讨论区 】
发信人: windwine (fly), 信区: Statistics
标 题: 今天的capital one statistician面试细节及抱怨，呵呵
关键字: capital one,statistician
发信站: BBS 未名空间站 (Thu Dec 2 16:29:52 2010, 美东)
今天去面试statistician,背景刚毕业的master，各个东西都懂一点但是都没有实战经
验。这边的主要是auto和Ｍｏｒｇａｇｅ的，然后也在建立CRM.
第一个是behavior，是里面的一个统计phd来问的:问过去的project有没有去问别人来
学习东西一起合作的经历，并阐述。文过去项目里面有没有需要学习新的方法来搞定项
目，并阐述过程。然后就随便扯了一下简历上面的一个如何predict yield curve 的
time series model问题。
第二个是case study,一个犹太人来问的。现在有信用卡，平均balance B,利息x,还有
平均的purchase amount S,然后对于purch... 阅读全帖 |
|
s*****e
发帖数: 21415 | 49 好了，前面的贴被大家说成没技术含量，现在有技术含量的来了。这篇
就说说如何跑赢大盘——alpha篇。
一般来说，投资的回报可以分解出来一个基本的factor, 就是大盘，或
者叫做market return。
R = Rm * Beta + Alpha
其中Beta通过least square fitting得到。如果是股票，beta也可以直接
从google，yahoo上查到。
基本上来说，几乎所有的投资都和大盘correlated，如果你的投资和大盘
基本无correlation，而且你这几年都在赚钱的话，那么恭喜你，你属于
典型有alpha的人群，你不应该是这篇小文的读者，你应该自己去开一个
fund。。。 呵呵
一个投资回报的时间序列，如果能够在扣除大盘correlated component之
后还有剩余正回报，那么这个投资strategy就是有alpha的。
注意这里说的不仅仅是简单的跑赢大盘，有可能你抓了一把小盘股，都是
high beta，大盘狂涨的时候你赚的比大盘多的多，于是认为自己“跑赢大
盘”，其实不一定，一算alpha可能是负的。这样的投资组合如果大... 阅读全帖 |
|
V*********n
发帖数: 198 | 50 association 更加broad, correlation 一般指linear correlation，但是可能会有
quadratic correlation，association可以泛指一般的关联，而causation就是指因果
关系了。 |
|
