l**********1
发帖数: 5204 | 1 LZ can try
stepwise selection method:
1
先單獨計算各X變項的p-value
2
逐一選出p-value<.05
3
加入model
4
進入model後再檢查ㄧ次p-value
5
Entry prob=0.05 or more
6
Removal prob=0.05 or less
cited from
one tw online ppt slide show:
//www.docstoc.com/docs/23426602/Essentials-of-Writing-Biomedical-Research-
Papers-Results-and
original
please go to
//www.stanford.edu/~kcobb/courses/hrp258/index.html
then click this one:
April 25: Statistical Inference I: CLT, confidence intervals, p-values
got that 80s page PPT file.
总的 |
|
s******9
发帖数: 283 | 2 http://www.sciencemag.org/content/early/2013/03/06/science.1236
The Brain Activity Map
A. Paul Alivisatos1,*,
Miyoung Chun2,
George M. Church3,
Karl Deisseroth4,
John P. Donoghue5,
Ralph J. Greenspan6,
Paul L. McEuen7,
Michael L. Roukes8,
Terrence J. Sejnowski9,*,
Paul S. Weiss10,
Rafael Yuste11,*
Abstract
Researchers propose building technologies to enable comprehensive mapping
and control of neural circuit activity.
Neuroscientists have made impressive advances in understanding the
microscale ... 阅读全帖 |
|
m******g
发帖数: 467 | 3 呼呼，下一个post终于要进入2014年了，感觉我写的内容越来越少了~~~
PNAS 13.12.24
稍微了解一下：
1. Mutational effects on stability are largely conserved during protein
evolution
Influenza nucleoprotein
Measure mutational effects on stability
In diverged homologs
Largely conserved
Simulated protein evolution is similar
小知识：
1. Genetic mapping of adaptation reveals fitness tradeoffs in Arabidopsis
thaliana
2. Visible light alters yeast metabolic rhythms by inhibiting respiration
3. Stepwise artificial evolution of a plant disease r... 阅读全帖 |
|
m******g
发帖数: 467 | 4 呼呼，下一个post终于要进入2014年了，感觉我写的内容越来越少了~~~
PNAS 13.12.24
稍微了解一下：
1. Mutational effects on stability are largely conserved during protein
evolution
Influenza nucleoprotein
Measure mutational effects on stability
In diverged homologs
Largely conserved
Simulated protein evolution is similar
小知识：
1. Genetic mapping of adaptation reveals fitness tradeoffs in Arabidopsis
thaliana
2. Visible light alters yeast metabolic rhythms by inhibiting respiration
3. Stepwise artificial evolution of a plant disease r... 阅读全帖 |
|
m******u
发帖数: 12400 | 5 the one stepwisely separate water-soluble,alcohol-soluble and alcohol-reduce
reagent soluble proteins. |
|
m******u
发帖数: 12400 | 6 the one stepwisely separate water-soluble,alcohol-soluble and alcohol-reduce
reagent soluble proteins. |
|
W*****o
发帖数: 1780 | 7 I am looking for a bench instruction using digitonin for mitoplast
preparation. Specifically, does digitonin need to be purified freshly each
time? What mechanical force is coupled to digitonin treatment, dounce
homogenisation, stir, or else? Appreciate if I can have a stepwise working
protocol. |
|
s******s
发帖数: 13035 | 8 不太明白这个均值为1有啥用途，可能是让你提到有些algorithm
要把predictor normalize吧
没做过biomarker, 不过这题目不是让你建模，而是让你找subset.
找subset要么自动找，要么手动stepwise找。前者可以用用lasso
啥的，后者就是把p-value大的，或者information gain多的predictor
一个一个加回去，直到predictor power(比如AUC)不增加为止。当
然最后要一个binary classifier, 那就在外面wrap一个logistic regression
就行了，比如logistic lasso.
机器学习玩的不多，有了feature selection，后面搞点naive bayes
或者decision tree就好了。random forest, neuron networks啥的不说
不管feature selection, 另外这些花样的结果比较难解释，药厂一般
不会搞这些吧。一般偏clinical side的都要容易解释.
现有
得范
达值 |
|
r****y
发帖数: 1437 | 9 just define a stiff stepwise function, then add to your
original cost function. |
|
|
i***c
发帖数: 220 | 11 这些都没戏的，必然损失大部分能量。dc dc都是分析稳态的，这个从0到v的过程不是
稳态。串啥电感二极管，stepwise，用电流源等的都没用。
go and search BUCK CONVERTER
or you may think about multi-source(i.e. use v/2 source charge first, then
use v source), multi-time, charge-pump, charge cap in parallel , charge cap
in serial...... |
|
|
|
i***c
发帖数: 220 | 14 这些都没戏的，必然损失大部分能量。dc dc都是分析稳态的，这个从0到v的过程不是
稳态。串啥电感二极管，stepwise，用电流源等的都没用。
go and search BUCK CONVERTER
or you may think about multi-source(i.e. use v/2 source charge first, then
use v source), multi-time, charge-pump, charge cap in parallel , charge cap
in serial...... |
|
|
w*********2
发帖数: 400 | 16 实在着急这里做不出来了,着急着今年毕业可是这部分作的差 .
现在的问题是 :
我用stepwise和GLR分别做出来air 污染的模型,模型马马虎虎,R2有0.7,也有差的0.3,
无所谓了,我用好的.
可是当我找其他的实验数据验证的时候,差别很大了.比如,实验数据是12,可是模型出来
的是1000.
怎么办呢?
请达人给点建议,虽然我知道这些模型都是骗人的,但是还是想能混过毕业阿. |
|
S*****n
发帖数: 6055 | 17 Stepwise Growth of a Single Polymer Chain
J. AM. CHEM. SOC. 2005, 127, 10462-10463 |
|
s*****1
发帖数: 273 | 18 如果大家有iphone，ipad并不是必须。但又大又超清晰ipad3的屏幕带来的享受还是很
值的。
我自己买了一个最便宜的ipad3 （$499）.
从两年前开始，一路ipad1， 2 使用下来，感觉很好。苹果的软件选择真是别家没得比
的。
日常用途
Email， 上网， 用dropbox, evernote, CHM+, Kindle 看资料，viber, google voice
, skype 交流
, USA today, ABC player, podcast， nietflix， TED, this American life 和 60
minutes, 随
时可以听/看到， 用零碎时间了解美国文化和学习语言。
对我而言，相对laptop和smart phone，更喜欢IPAD3集下面三点于一身。
1. 便携
2. 屏幕够大超清晰
3. 随时启动
==
下面是我在用的apps.
Merriam-Webster dictionary
Atomic web (Browser for checking ERAS website)
drug pronunciations
medic... 阅读全帖 |
|
x********8
发帖数: 167 | 19 前面几个SEMINAR讨论了很多单壁管近红外荧光受pH值和分子的影响，单壁管能发光，
而且发光行为恰好验证了Dresselhaus早年对于单壁管电子结构的计算，确实是一个很
好的实验研究方向。但是分子如果和单壁管共价键结合的话，近红外发光常常被完全淬
灭掉，不是一个好现象，前面提到的做敏感sensor应用就是用到biotin和单壁管的非共
价键结合的性质，这样荧光可以恢复过来。最近又看到一篇Science杂志上最新的关于
单壁管微区荧光研究，用到了单分子光谱技术，观察到了在单根单壁管在长为670nm区
域的荧光动力学变化。这个手段确实很厉害，居然把单个激子对的非辐射复合过程给拍
出来了。
Stepwise Quenching of Exciton Fluorescence in Carbon Nanotubes by Single-
Molecule Reactions
Laurent Cognet,1,2* Dmitri A. Tsyboulski,2 John-David R. Rocha,2 Condell D.
Doyle,2 James M. Tour,2 R. Bruce |
|
l******i
发帖数: 1404 | 20 得出的值方向不对，说明model本身有问题。
选择predictor：如果predictor数目多的话（注意multicollinearity），建议用
stepwise model selection；predictor数目少的话，用all subset search.
在所有感兴趣的model里用R_adjusted_square结合C(p)和PRESS(p)判断。也可以结合
AIC, BIC, SIC。判据越多越好。
不过唯一的最好办法：get much more data, need sample size to be large enough.
这样asymtotic理论下的所有test就有意义了。也就不存在你说的Rsquare一会儿大一会儿小的问题了。
结果实在差的很远，不整regression，用multifactor ANOVA Model
再不行，就用nonparametric approach呗，例如bootstrapping |
|
m*********g
发帖数: 646 | 21 not very sure about your question, but try to google "aic stepwise nonlinear
regression model" |
|
m*********g
发帖数: 646 | 22 then, i think the nonlinear version "aic stepwise" could work. you can just
google some basic techniques, some include codes in R or in SAS
or, you can write your own codes based on some criteria score.
不大 |
|
y*******o
发帖数: 6632 | 23 stepwise and calculated information lose weighted by variable number.
AIC, BIC.
不大 |
|
a**a
发帖数: 32 | 24 Stepwise, AIC, BIC, Bayesian Score, LASSO, BOOST. |
|
w**********y
发帖数: 1691 | 25 These interviewers are interesting. Given your background and their
questions, can't they even tell out the mis-match from the very beginning?
Ridge can't do variable selection. So p predictors in, p predictors out;
LASSO does. Different penalty scale gives different numbers of selected
predictors.
Mixed of ridge and LASSO, elastic net, also does. Others include SCAD, SIS,
and the old school but wrong method - stepwise.
Dimension reduction is a different topic with variable selection. This
direc... 阅读全帖 |
|
m**********4
发帖数: 774 | 26 大牛能否讲讲为啥old school的stepwise regression is wrong？我一直以为这种
greedy methods在现实中满实用的呢。
, |
|
w**********y
发帖数: 1691 | 27 I meant it has theoretical flaws. To list top 3 in my mind: wrong F-test at
each step, wrong P-value, and wrong degree of freedom.
See here for more:
http://stats.stackexchange.com/questions/20836/algorithms-for-a
In situation with severes multicollinearity, both stepwise and lasso will
perform badly. |
|
G******2
发帖数: 579 | 28 他们没有report p value 和t statistics .我还在想是我不懂的原因，一般来说我先
看p value. 不过我可以要求report
R square 看起来是model 的。
给我的sad code 打不丶开。我家电脑没有sas.
说是用的Bayesian schawz criteria selection. 跟我平常用的stepwise 和backwards
selection 不一样。不懂呀。
最小的SBC 和最大的RSQ的variable 被淘汰了，说intuitive judgement 不make
sense 是个dummy variable
用了两个variable 说 SBC 最小值，然后rsq 也最大。然后multivaraite 用了这两个
variable 一起，beta 都不大，没有report p value。f statistic 也没有给我。
我真希望当初读书时多念点Stats 。这都是phd 写的report。我看得稀里糊涂的。可是
平常看文献这些p value是一定要report的。
我是不是应该问这些phd要求数据支持他们选择... 阅读全帖 |
|
w**********y
发帖数: 1691 | 29 你问的问题太弱跟你讲起来太麻烦估计要给你补好多101 102的课
简言之，给你东西的人听起来很懂他们的描述非常make sense。不懂的直接请他们教你
吧。
forward selection, stepwise
都是最老的做regression 的方法而且都是被证明了是错误的至少只是approximation.
展开讲太多知识了。你如果常用这些去统计系修两门相关课程吧 |
|
r*******g
发帖数: 453 | 30 请教一下regression的话不用forward, stepwise 该用什么呢？
. |
|
p********a
发帖数: 5352 | 31 ☆─────────────────────────────────────☆
lilyrong (ATLAS) 于 (Wed Feb 21 12:39:29 2007) 提到:
hello, could somebody please help me out? Thanks so much!
I did logistic regression model which try to predict disease using 10
variables. I have 700 (pts+ normal controls).
when I use individual variable to predict disease, both variable 1 and 2
have odds ratio >999.99.
95% CL are from >999.99 -999.99 for variable 2, and 567.9 to >999.99 for
variable 1.
When I combine those and use stepwised logistic |
|
p********a
发帖数: 5352 | 32 ☆─────────────────────────────────────☆
waterpuma (waterpuma) 于 (Wed Jan 2 14:15:08 2008) 提到:
This is a problem form onsite interview.
There are more than 100 variables. Based on those variables, how to
regression a model for one depend variable? We need use all those variables
or pick up some variables? If we need choose some variables, which is the
best way to choose it, let the option of stepwise do it or we should use
factor and princomp do it?
Thanks for any response!
☆─────────────── |
|
w********r
发帖数: 253 | 33 你的意思是还是全部放进去model
然后用stepwise之类的步骤来select？
我觉得结果会很trick。。。。
regularization是什么？ |
|
h***x
发帖数: 586 | 34 你可以看看data mining cookbook p106-108, MC does not have any adverse
effects on a predictive model.
Stepwise一般不会有预期的那么好。把他们都放到model里，fit一下，看看chi square
，取值大的那个变量。
如果correlation coefficient很大，try PCA and factor analysis to check if they
can be combined into one variable. |
|
p********a
发帖数: 5352 | 35 ☆─────────────────────────────────────☆
xy645598 (Heather) 于 (Wed May 14 11:17:00 2008) 提到:
Statistical Analyst/SAS Programmer职位。上周电面HR，谈的比较开心，就问些简历
和behaviour question.今天面试的是tech 问题，project & basic sas都还好，结束
前20分钟被问到MACRO,平常用的不多，option有点不熟，感觉回答的不好。最汗的是，
问知道ICD-9吗，我说知道，然后他说choke的ICD-9 code是什么。鬼知道，我只好说记
不住了。完了，Move on.
☆─────────────────────────────────────☆
xy645598 (Heather) 于 (Wed May 14 11:45:48 2008) 提到:
刚才忘了一个问题，STEPWISE procedure 的优缺点，我觉得自己回答的不好，在这里
请教一下。谢谢
☆───────────────────── |
|
t******u
发帖数: 1 | 36 据说orthoreg可以算数据之间有关联的线性回归，但因变量可能不是二值变量吧...用
一般的回归如reg中的stepwise好像也可以，但不太确定是不是您提到的multivariate
regression... |
|
T******r
发帖数: 265 | 37 1) pca中用什么方法来选择应该用几个factor呢？
2) 可不可以用pca, 然后再用stepwise 或者其他方法选factor.还是最好用前几个
factor
3) 用pca做forecast.如何避免in-sample overfitting的问题呢？
4) 有没有什么关于pca的教程呢 |
|
T******r
发帖数: 265 | 38 我的问题是
1. 我有10，000 explanatory variables 但只有1000 records
2. 我想用pca reduce dimension
3. 然后用factor forecast dependant variable.
4. 我不知道我对这个variation的理解对不对，我想是从大到小选择前几个factors.但
是这样的forecast accuracy 不太好 但如果我用stepwise 选择 factors的话，我觉得
有in-sample overfitting的问题。 |
|
T******r
发帖数: 265 | 39 PCA出来的factor都independent了还要stepwise干什么？
我是想选出和dependant variable最相关的factors.我的理解有错吗？
这个independant 指得是orthogonal 吗？
为什么需要正定呢？
还有哪位高手看看我9楼的问题呢？ |
|
b*******y
发帖数: 30 | 40 use stepwise, which is more popular nowadays. |
|
r****e
发帖数: 1 | 41 我只学了一点儿统计，所以问的问题可能很傻，先谢谢大家了。
现在我有两组数据，一组是原材料的性能，一组是产品的性能。每个lot原材料测试2个
数据，产品测试5个数据。一共有100多个lot。然后我还加上了可能的其它factor，测
试人，测试仪器，什么的。用stepwise，原材料的性能不是main factor。我觉得我一
定是哪里做错了，请问还有什么方法判断两组数据是否相关？ |
|
j*********9
发帖数: 161 | 42 需要做一个multiple regression model的project，有8个变量。可是具体步骤
我还不是很清楚（羞愧。。），我只知道要画scatter plot，stepwise selection，还
要分析interaction。
可是最好的步骤应该是怎样的呢？
希望高手指点一下，多谢！ |
|
h******e
发帖数: 1791 | 43 用stepwise的方法选择变量，得到两个A和B,都是continuous的，两者之间的
interaction被排除了，如下：
Parameter DF Estimate SE Wald Pr>ChiSq
A 1 0.1297 0.0435 8.9076 0.0028
B 1 -0.3586 0.1732 4.2843 0.0385
可见两者的作用是相反的。
但如果单独用A或B做logistic model，得到：
A的estimate是0.0565，B的estimate是0.1665，两者的作用是同向的。如果单独对二者
做linear regression的话，发现存在正相关性。
请问该如何解释logistic regression的结果。谢谢。 |
|
|
|
g*****c
发帖数: 4378 | 46 two comments
1) for the statistical analysis you listed, make sure
you really understand the detail of each. you claimed
you had two years experience, got to be familiar with
them. for example, people may ask you teh mathematical
formula for logistic regression, the sas procedure,
stepwise option, interpretation of the model etc.
by the way those you listed are not complex analyses
2) you may not want to list all the other softwares depends
on what position you apply for.goo |
|
h******e
发帖数: 1791 | 47 前辈，有这么一个logistic regression的问题，请帮帮忙：
In this logistic regression (real clinical data) with 22 severe diseased and
25 less diseased, I used stepwise method to select covariates with SAS
procedure logistic. The objective is to examine variables which could cause
disease progression. I finally got two: A and B. Both are continuous. Their
interaction A*B was ruled out by the SAS procedure.
Parameter DF Estimate SE Wald Pr>ChiSq
A 1 0.1297 0.0435 |
|
l***a
发帖数: 12410 | 48 this algorithm sounds interesting... what's it called? |
|
f****7
发帖数: 398 | 49 谢谢 minner! 刚才看了Proc reg的东西，感觉很有帮助。有一个问题是sas proc reg
对predictor的选择是基于R-square的基础上，我可能比较偏向基于correlation的基础
上，不知道有没有什么方法可以将两者联系起来？ |
|
s*r
发帖数: 2757 | 50 proc reg won't do the addition that you have described.
you did not realize that libra was a little sarcastic ?
reg |
|
