c**e
发帖数: 2558 | 1 【 以下文字转载自 Memory 讨论区 】
发信人: cute (小可爱), 信区: Memory
标 题: [M] The Corrs -What Can I Do
发信站: BBS 未名空间站 (Wed Jul 4 12:53:22 2007), 站内
http://www.youtube.com/watch?v=6wTHjYcsdQ0[
Music by: the corrs lyrics by: andrea corr
I havent slept at all in days
Its been so long since weve talked
And I have been here many times
I havent slept at all in days
I just dont know what Im doing wrong
Its been so long since weve talked
And I have been here many times
I just dont know what Im doing wrong
What can I do |
|
R******d
发帖数: 1436 | 2 我有一个很大的数据集，要计算25000个变量和将近3000个变量之间的相关性。
结果出来，发现有很多空行没有相关系数,比如L19, L33这样的。如果把为空的数据点
（L19, L33）单独拿出来，则可以算出结果。请问这是什么问题？
谢谢了。
data mydata;
infile "/dir/file" firstobs=2 lrecl=2000000;
input C1-C25000 L1-L3000;
run;
proc corr data=mydata outp=corr(where=(_NAME_ ne "") drop=_TYPE_) noprint;
var C1-C25000;
with L1-L3000;
run;
proc export data=corr
outfile="/dir/out"
dbms=tab replace;
run; |
|
t*******n
发帖数: 1724 | 3 Corrs 一家子都很牛
另外跟 Josh 合作的乐队/歌手组合是 The Corrs
★ Sent from iPhone App: iReader Mitbbs 7.56 - iPad Lite |
|
p*****o
发帖数: 543 | 4 ODS OUTPUT PearsonCorr = MyOut;
PROC CORR DATA = INPUT;
VAR A B C;
WITH D;
BY E F;
RUN;
ODS OUTPUT CLOSE;
上面这个CODE中BY STATEMENT会给出所有的E和F的组合，并把结果存到MYOUT中，可是
我还想得到只按E分组，跟只按F分组。我当然可以写三次上面的CODE，把BY STATEMENT
分别改成BY E和BY F；可是我想知道，PROC CORR中有没有办法可以在一个PROCEDURE里
完成我要的东西的？ |
|
A*****1
发帖数: 1738 | 5 Yeah
今儿
不吃不睡了
实在闲的发慌
上房揭瓦一回
借光了您哪
一胡同串子
也撒丫子跑出来
过过场儿
越活越抽抽儿
所以
不知道什么是肝儿颤
Yeah
So Young - The Corrs
Yeah Yeah Yeah Yeah Yeah
We are taking it easy
Bright and breezy, yeah
We are living it up
Just fine and dandy, yeah
We are caught in a haze
On these lazy summer days
We’re spending all of our nights just
A-laughing and kissing, yeah
And it really doesn't matter that we don't eat
And it really doesn't matter if we never sleep
No it really doesn't matter, really doesn't matter at all
Coz we are so y... 阅读全帖 |
|
|
t*******r
发帖数: 22634 | 7 另外跟 Josh 合作的乐队/歌手组合是 The Corrs |
|
r*****t
发帖数: 286 | 8 ☆─────────────────────────────────────☆
artwork (嘿嘿) 于 (Tue Jan 30 16:47:40 2007) 提到:
就是3个random variable之间的corr可能与不可能
得题， xiaxia
☆─────────────────────────────────────☆
loveleader (hatefree) 于 (Tue Jan 30 17:35:11 2007) 提到:
Let a, b, and c are correlations between three random veriables, then
a^2 + b^2 + c^2 - 2 a * b * c <= 1
You can find this in John Hull's book (edition 5) at page 388's foot notes.
For more details, you may refer matrix analysis book.
☆──────────────────────────── |
|
p*****o
发帖数: 543 | 9 弱弱地问一句。。。PROC CORR中也有CLASS STATEMENT么。。。 |
|
o****o
发帖数: 8077 | 10 木有，我错了。
一个办法是这样搞
DATA NEWV/VIEW=NEWV;
SET OLD(IN=_1) OLD(IN=_2) OLD(IN=_3);
IF _1 THEN NEWID=ID1;
ELSE IF _2 THEN NEWID=ID2;
ELSE IF _3 THEN NEWID=CATS(ID1, ID2);
RUN;
PROC CORR DATA=NEWV OUT=.....;
BY NEWID;
VAR &SELECTEDVARS;
RUN; |
|
J******m
发帖数: 97 | 11 请问你做corr 的目的是什么？ 是Prepare inputs 好为做模型做准备么？
你的问题不知道怎么回答，你看SAS那本书：叫predictive modeling using logistic
regression, 选变量的顺序一般是先做cluster吧， 经过这一步就能去掉很多变量，第
二部才做相关性分析。回答如果不对请见谅！ |
|
d******e
发帖数: 17 | 12 可以，直接regression, correlation就行了；
但结论仅对此人成立,且不一定和group-pooled correlation相近；
Denote Z= Individual No
因为你估计的是 corr(X1,X2 | Z=1),
Need corr(X1,X2 | Z=i)=corr(X1,X2 | Z=j), for any i, j
且 corr(X1,X2 | Z=i)！=corr(X1,X2）
所以结论是你的estimator coverge to corr(X1,X2 | Z=1) 且还需假设 n1 ->\infty.
------------------------------------------
1、即使所有人都相同的 correlation within individual, group correlation is
not equal to correlation within individuals. 这和把一个longitudinal data进行
pool算一样；
2. 需要假设individual correlation equa... 阅读全帖 |
|
m****m
发帖数: 4741 | 13 2009.9.11
说什么好呢？
The Corrs - Unplugged
http://www.amazon.com/Unplugged-Corrs/dp/B00002R0K8
那些年听欧美流行音乐， The Corrs是我钟爱的乐队之一
欧洲的流行音乐比美国的更注重旋律。
The Corrs的音乐琅琅上口，几乎每首都很好听。
兄妹4人能唱能弹能拉，主唱小妹妹Andrea慵懒的声音和她的容貌一样迷人。
这张Unplugged里收集了The Corrs前几张唱片里的经典，最后那首翻唱REM的"everybod
y hurts"相当精彩
only when I sleep
Radio
还有翻唱Paul McCartney的no frontier |
|
k**y
发帖数: 920 | 14 Corr(X1+X2,Y1+Y2)可以等于 [-1,1]上任何数，对吗？
用老周的方法，画图，X1,Y1是夹角小于arccos0.9的两个向量，长度任意
X2,Y2是夹角小于arccos0.9的两个向量，长度任意
问向量X1+X2和Y1+Y2的夹角。
画个图就可以发现，这夹角可以[0,180]
比如假设corr(x1,y1)=corr(x2,y2)
那么当x2=y1, y2=x1的时候，corr(x1+x2,y1+y2)=1
当x2=-y1, y2=-x1的时候，corr(x1+x2,y1+y2)=-1 |
|
k**y
发帖数: 920 | 15 Corr(X1+X2,Y1+Y2)可以等于 [-1,1]上任何数，对吗？
用老周的方法，画图，X1,Y1是夹角小于arccos0.9的两个向量，长度任意
X2,Y2是夹角小于arccos0.9的两个向量，长度任意
问向量X1+X2和Y1+Y2的夹角。
画个图就可以发现，这夹角可以[0,180]
比如假设corr(x1,y1)=corr(x2,y2)
那么当x2=y1, y2=x1的时候，corr(x1+x2,y1+y2)=1
当x2=-y1, y2=-x1的时候，corr(x1+x2,y1+y2)=-1 |
|
D******6
发帖数: 6211 | 16 嗯，一年夏天一个朋友送我四张优惠音乐演出票，我根本不知道the corrs，反正是优
惠票，不去白不去，集合了4个人去了。在英国待过的朋友听过the corrs，告诉我是个
好乐队。回来我们还讨论为什么叫the corrs，后来才知道他们是姓corr的一家子，呵
呵。
当时那个演出前面还有一场是另外一个女歌手，我们很不喜欢她的台风，那个女歌手都
忘了名字了，估计是个不怎么样的。 |
|
B*********h
发帖数: 800 | 17 ☆─────────────────────────────────────☆
pangzi (艾玛尼) 于 (Fri Dec 1 00:29:36 2006) 提到:
x,y,z random variable,
corr(x,y)=corr(y,z)=0.9
then can corr(x,z)=0.5?
corr is the correlation
☆─────────────────────────────────────☆
muying (心情娃娃～※～～～～) 于 (Fri Dec 1 00:41:51 2006) 提到:
no
☆─────────────────────────────────────☆
pangzi (艾玛尼) 于 (Fri Dec 1 00:48:29 2006) 提到:
why?
☆─────────────────────────────────────☆
Angel (On leave) 于 (Fri Dec 1 00:55:00 2006) 提到:
minimum correlation i |
|
s****i
发帖数: 475 | 18 可以用corr(x1,y), corr(x2,y), 和corr(x1,x2)的定义来算出corr(y,y_hat)么，这样
的话好像这个未知数太多，比如b1_hat 和b2_hat |
|
w***z
发帖数: 28 | 19 just woudering: corr(w,h) and corr(w, p) are just two numbers, how can and
possible to do statistical test for 2 numbers?
have a sample of patients, each has variable
weight (w), height (h) and blood pressure (p).
How to compare corr(w,h) and corr(w, p) in SAS?
What is the statistical test?
Thanks a lot, |
|
l******o
发帖数: 67 | 20 Corr(X1,Y)=p1,
Corr(X2,Y)=p2,
what is the max value of
Corr(a*X1+b*X2,Y) for any a,b.
我自己推了下，发现 没用到 Corr(X1, X2)， 感觉肯定不对。 |
|
h**********0
发帖数: 1453 | 21 不是所有东西是绝对的
希特勒一样有一个美军的侄子
从统计上说，corr能达到1才是一定
但是人的性格绝大部分是不可能跟周遭的环境因素的corr都能达到1
但并不代表没有联系
有兴趣的自己去抽样，去做调查
我相信楼主说的那些corr都会在0.5以上，cofidence在95%
你非得把outlier拿出来说，那其实没啥可说的 |
|
h**********0
发帖数: 1453 | 22 你不用关心我的计划
我还可以给你一个数据
根据我们的统计资料
有孩子的女性跟没孩子的女性对比
没孩子的女性中风的可能性更大，而且呈一定的正比关系
corr＝0.53，其它心血管疾病的概率也会增加
corr＝0.37，自杀的比例更高一些
corr=0.51，总之是那么一回事，你尽管high去吧 |
|
t*****t
发帖数: 244 | 23 如果已知这两个时间序列都各自由某个process/model来产生，那correlation理论上也
是知道的，
具体取决于underlying models。
如果你就想通过测量实时的样本数据来动态测量correlation，那只有用expanding
window或者
rolling window去动态测量correlaton。当然怎么测量取决于你的选择，sample corr,
robust corr, 或者要不要考虑time varying volatility，等等。
correlation是一个抽象概念，并不存在一个准确值。当然你可以用forward realized
corr作为
benchmark来衡量你的算法的优略。 |
|
l**********1
发帖数: 5204 | 24 别轻言 转进 from cns and N sister higher IF jounrlas 关键是倒数第一 或 第
二位
corres or co-corres author/s 若非 cns过去5〜6年 平均 每间隔一年 必有
一篇 最后/通讯作或共通讯作的 记录之大牛的话 够呛
此cns last/corres author/s 位置是当年的罗马元老院 类似 的 学术政治
authorship
当然没有以上历史记录的组 除非投稿的是 可以发到NEJM or Lancet 上去的临床结果
降格投稿N Med 才可能中。
http://www.mitbbs.com/article_t/Biology/31782095.html
所以非 iSun (I sun) 不能也 实没投到大牛门下 或至少挂名 合作署名 的共通讯
的大牛门路也
ps:
比如 Jessell TM offer you iSun (I sun) as his one PD then 2 Ys later,
Cell N or N sisiter/... 阅读全帖 |
|
J*******g
发帖数: 267 | 25 I think at most an upper bound and a lower bound could be found for corr(x,
z). Looking at the correlation matrix of the random vector (x, y, z), which
would be a valid correlation matrix iff it is positive semidefinite. Note
that the only missing number in the correlation matrix is corr(x, z). Use
the positive semidefinite property, you can derive the bounds on corr(x, z). |
|
c*******o
发帖数: 1722 | 26 corr will definitely not do this type of time series. there is obvious
patten/trend in
the time series, which you want to get rid of before any statistics test.
i would suggest Fourier analysis first and do corr for the residues. or use
ar/garch
first and corr afterward. |
|
h******g
发帖数: 69 | 27 面过他们的intern,不过是Quant
第一轮店面， 问一些纯C的问题，例如
char s[10];
char *x = malloc(strlen(s));
Is anything wrong?
Write a C function to compare two strings (case independent)
第二轮店面，问一些element数学题
What is vector space?
What is Brownian Motion?
Rank of A*B, where A, B are two matrix
Three r.v. X, Y, Z. corr(X,Y)=0.6, corr(Y,Z)=0.8, What is range of corr(X,Z)
List some numerical analysis algorithms
What is R^2 in linear regression?
Suppose y=a+bx, we now have x1,...,xn and y1,...,yn. How to estimat |
|
l***m
发帖数: 920 | 28 如果corr(x1,x2)=-1,也就是说x2 = - a * x1, a>0.
请问如何找一个x3,使得corr(x1,x3)=corr(x2,x3)=-1呢???
}{ |
|
t***l
发帖数: 3644 | 29 例子2中的B_t都不是Martingale，怎么会是BM呢？
而且两个corr都差乘个t
我猜楼主说的是corr一定为常数乘以t吗？
我来给个非常数的例子:
假设W_t和B_t是两个独立的BM，又假设a(t)是个取值0到1之间的函数，让b(t)=sqrt{1-
a(t)^2}。那么 a(t)W_t + b(t)B_t就是一个BM，它和W_t之间的corr是a(t)t |
|
h***t
发帖数: 2540 | 30 Let's say you have 3 random variables X, Y, Z with corr(X,Y)=0.70, corr(X,Z)
=0.80. What is the minimum value for corr(Y,Z)？
Thanks |
|
z*******g
发帖数: 18 | 31 method 1. correlation between two variable can be expressed as the angle
between them, so corr(X,Y) = cos(x1)=0.7, corr(x,Z)=cos(x2)=0.8,
so the minimum value of corr(Y,Z) = cos(x1+x2);
method 2. using the property that the covariance matrix is positive defined.
Z) |
|
l******i
发帖数: 1404 | 32 是那个年轻的概率PhD问你的这题吧？
我也贡献一题以前在SIG onsite被问过的：
已知Corr(X1,Y1) > 0.9, Corr(X2, Y2) > 0.9.
What can you say about Corr(X1+X2,Y1+Y2)?
y, |
|
E****t
发帖数: 6 | 33 可不可以这么考虑：
上限是 1，取 X1 = Y1 = X2 = Y2.
下限是 －1：
比如Z ～N（0，1）， X1 ＝ Y1 ＝ Z， X2 = -2Z, Y2 = -0.5Z,
corr(X1, Y1) = 1, corr(X2, Y2 ) =1, corr(X1+X2, Y1+Y2) = -1.
所以没有任何限制？ |
|
l******i
发帖数: 1404 | 34 是那个年轻的概率PhD问你的这题吧？
我也贡献一题以前在SIG onsite被问过的：
已知Corr(X1,Y1) > 0.9, Corr(X2, Y2) > 0.9.
What can you say about Corr(X1+X2,Y1+Y2)?
y, |
|
E****t
发帖数: 6 | 35 可不可以这么考虑：
上限是 1，取 X1 = Y1 = X2 = Y2.
下限是 －1：
比如Z ～N（0，1）， X1 ＝ Y1 ＝ Z， X2 = -2Z, Y2 = -0.5Z,
corr(X1, Y1) = 1, corr(X2, Y2 ) =1, corr(X1+X2, Y1+Y2) = -1.
所以没有任何限制？ |
|
A*********u
发帖数: 8976 | 36 用 proc corr
I have a sample of patients, each has variable
weight (w), height (h) and blood pressure (p).
How to compare corr(w,h) and corr(w, p) in SAS?
What is the statistical test?
Thanks a lot, |
|
c**********e
发帖数: 2007 | 37 Thanks. But my problem is to test if two correlations
are equal. For example, proc corr gives corr(x,y)=0.6
and corr(x,z)=0.7. Are the two correlations signicantly
different? (Apparently, sample size makes a difference.) |
|
n*****1
发帖数: 172 | 38 less than 0 but close to 0
小于0是明显的. 如果a和c都只有2个element, 那么corr是-1, 当分别在a和c后增加相
同的element的时候, 相当于a和c都
在增加corr是0的element, 那么总的corr是在-1和0之间, 当后面增加的相同的element
越多, 越接近于0.
sequences |
|
f******a
发帖数: 129 | 39 后面增加的element,corr可以是0，也可以是1吧，因为他们的方差也是0啊，没法确定
他们的corr。因为分子分母都是0. 所以我认为应该选择3. corr小于0是肯定的，当时
能否接近于0不一定，要看分母的是否也接近于0.
element |
|
D*********2
发帖数: 535 | 40 Yes, it works when I use
> dat.corr <- read.csv(file="xxx.csv", colClasses=rep("factor", 4), T)
However, I did not notice, even use read.csv, it gave me "factor". Why?
I simply did import to .csv in SAS.
> dat.corr <- read.csv(file="xxx.csv", T)
> sapply(dat.corr, class)
Y X1 X2 X3
"integer" "factor" "factor" "factor"
Y is the one should be factor, say, "007". X1:X3 should be numeric.
Thanks again! |
|
f******h
发帖数: 46 | 41 谢谢，刚刚上面回了一贴，说处理过multicollinearity以后的变量们发现并不是最好
的pool。。。我觉得是我multicollinearity处理方式不对头。我是在每一步去掉VIF最
大的那个变量，我也注意到这样的方式，很容易导致把那些和dependent variable的
correlation最大的predictor都去掉了。。。很ft
我想试试你的方法，在每组corr很大的变量中保留那个univariate R^2最大的。但是这
里也有问题：１）因为变量非常多，这种大corr的组合并不是mutually exclusive的，
就是说组和组的不同变量之间也很难避免一些corr很大，当然，这个可以考虑用
cluster analysis来交给sas解决；２）另一个问题是每组保留一个可靠吗？还是说在
经验上这样的做法是一种惯例？
correlated |
|
c**********8
发帖数: 38 | 42 问大家个统计题, (X,Y,Z) are three dimensional normal random variables, corr(
X,Y)>0, corr(X,Z)>0, can corr(Y,Z)<0 ? why? |
|
K*****2
发帖数: 9308 | 43 可以，相关矩阵比如说是
1.0 0.1 0.1
0.1 1.0 -0.1
0.1 -0.1 1.0
仍然正定。从三维独立正态出发，用这个矩阵就能构造出符合条件的X,Y,Z。
如果知道corr(X,Y)和corr(X,Z)的具体数值，那么corr(Y,Z)只能在一定的范围内变化
（为了保证相关矩阵正定）,这个范围可以从Y,Z|X的partial correlation处于-1和1之
间确定。 |
|
i**M
发帖数: 108 | 44 Let X_t, Y_t, and Z_t be three time series with known volatilities and we
know corr(X_t,Y_t) = p_t and corr(X_t, Z_t) = q_t.
Is there a way to find corr(Y_t, Z_t)?
Thanks |
|
e***e
发帖数: 3872 | 45 Fernando Pessoa
Poemas de Alberto Caeiro
O Tejo é mais belo que o rio que corre pela minha aldeia,
Mas o Tejo não é mais belo que o rio que corre pela minha aldeia
Porque o Tejo não é o rio que corre pela minha aldeia.
O Tejo tem grandes navios
E navega nele ainda,
Para aqueles que vêem em tudo o que lá não está,
A memória das naus.
O Tejo desce de Espanha
E o Tejo entra no mar em Portugal.
Toda a gente sabe isso.
Mas poucos sabem qual é... 阅读全帖 |
|
e***e
发帖数: 3872 | 46 Fernando Pessoa
Poemas de Alberto Caeiro
O Tejo é mais belo que o rio que corre pela minha aldeia,
Mas o Tejo não é mais belo que o rio que corre pela minha aldeia
Porque o Tejo não é o rio que corre pela minha aldeia.
O Tejo tem grandes navios
E navega nele ainda,
Para aqueles que vêem em tudo o que lá não está,
A memória das naus.
O Tejo desce de Espanha
E o Tejo entra no mar em Portugal.
Toda a gente sabe isso.
Mas poucos sabem qual é... 阅读全帖 |
|
T**********y
发帖数: 559 | 47 新AP，写篇review, 几乎全是我写的；
1. 让博后 first author, 我senior author/Corres
2. 找个领域内有影响的牛人，我first/corres, 牛senior; （牛人估计能找到）
有个牛人挂 估计好点的review journal简单些，但是我还是first author;
大家怎么看？ |
|
f******u
发帖数: 250 | 48 (导入数据)
proc import out=aapl
datafile="C:\Users\aapl.xlsx"
dbms=xlsx replace;
sheet="sheet1";
getnames=yes;
run;
proc import out=spy
datafile="C:\Users \spy.xlsx"
dbms=xlsx replace;
sheet="sheet1";
getnames=yes;
run;
(筛选数据)
data aapl;
set aapl(keep=date adj_close);
label adj_close="Aclose";
rename adj_close=Aclose;
run;
data spy;
set spy(keep=date adj_close);
label adj_close="Sclose";
rename adj_close=Sclose;
run;
(整理数据)
proc sort data=spy;by descending date;run;
proc sort data=aapl;by descendin... 阅读全帖 |
|
Y**b
发帖数: 73 | 49 现在8个审稿 离10次很近啊 很快就攒够了
[在 corr (corr) 的大作中提到：]
：计算化学方向
：文章 12
：........... |
|
G****a
发帖数: 10208 | 50 Locals were irate when the drug giant Pfizer closed its 70-hectare research
and development (R&D) facility in Ann Arbor, Michigan, in 2007. T-shirts
sporting the word 'Pfired' appeared on the streets; the governor called the
lay-offs a “punch to the gut”; and the state of Michigan pledged US$1
million to help the 2,100 displaced workers find new jobs.
The pharmaceutical industry has faced major upheaval in recent years, with a
disappointing drug pipeline, major revenue losses as patents expire o... 阅读全帖 |
|
