由买买提看人间百态

topics

全部话题 - 话题: bonferroni
首页 上页 1 2 (共2页)

发帖数: 1
1
来自主题: Biology版 - adjacent p和FDR的区别?
我不是很懂, 抛砖引玉
P-value 就是false的discovery (positive)的概率,用统计的话说就是,incorrect
rejection of a true null hypothesis (也就是你的 false positive)。 但是这个是
针对single test, 但是对于multiple test, 在用p-value就会造成过多的false
positives, 所以需要adjust, 最简单的就是Bonferroni, 直接除以the number of
tests, 但是这样的话会造成很多true positives 不能被发现, 降低了power. 于是,
BH 提出FDR(False discovery rate), 其实也是一种adjust p-value的方法 for
multiple tests, 是p-values 排序(step-up, step-down)的方法来找到threshold. 这
样能保证false positive 在一定level情况下, 找到尽可能多的true positives. 但
是BH证明时, se... 阅读全帖
y******r
发帖数: 34
2
来自主题: Computation版 - what's Bonferroni correction of p-value
who can tell me the formula? Thanks a lot!
n*********0
发帖数: 22
3
来自主题: Pharmacy版 - 请教关于multiplicity adjusted p-value
请问你说的multiplicity adjusted p-value是simultaneously testing, adjusted
family p-value吗?就是同时进行多项比较,控制family p-value小于一定值,而不是
控制单项p-value。
如果是的话,可以用sas软件计算。有三种procedures, Tukey, Scheffe,和
bonferroni。 具体的用法和code网上有,可以搜索到。
F********E
发帖数: 1025
4
Got it! Some guys provide a matlab script which performs the T-test with
Bonferroni Correction.
h******e
发帖数: 1791
5
来自主题: Statistics版 - 还是三组数据的问题。
如组a,b,c,其中a是独立的,但b和c是配对的,在这种情况下我可不可已直接用t-
test做两两比较,但用bonferroni的方法校正显著性水平?DaShagen同学,你的方法很好,但是这个数据来自临床,我不知道如何向医生解释。
h******e
发帖数: 1791
6
【 以下文字转载自 Biology 讨论区 】
发信人: hehehehe (入了大圣教,心中充满喜乐), 信区: Biology
标 题: multiple comparison方法的选择。
发信站: BBS 未名空间站 (Sun Feb 21 22:41:55 2010, 美东)
有什么根据吗?在什么情况下选择tukey,什么时候用bonferroni,什么时候用scheffe
?谢谢。
x***x
发帖数: 3401
7
我在实际应用(一般建模)中都用bonferroni
觉得除非特殊情况 不会用其他的
B****n
发帖数: 11290
8
来自主题: Statistics版 - 请教F test和t test区别?
Let's assume the significance level be 0.05.
You can definitely use 3 t-tests to make pairwise comparisons for 3 samples.
However, in order to keep the overall type I error (familywise error) still
being 0.05, you have to make the individual significance level smaller (
Bonferroni's correction is one option). This is so called multiple
comparison problem. Once you adjust this problem, the resulting t-tests has
familywise error equal to (or less than) 0.05, but the power is much less
than the F t
a****m
发帖数: 693
9
如果我想做两个group, control group(16 patients), treatment group (14
different patient given the treatment), 而且对每个patient做五个不同时间的测
量,如
果对于任何时间点的对照组和treatment,看有没有差异,用什么teset?
一般的t test with bonferroni correct是不是太严格了?
还有Fisher's LSD and Tukey's HSD? 好像很多人perfer Turkey's HSD.
还有什么repeated ANOVA,不知道谁有这方面经验?
3x
y*********e
发帖数: 298
10
来自主题: Statistics版 - 求本书
我刚刚接触统计这方面,有没有入门级介绍
False discovery rate
one way ANOVA
two way ANOVA
tukey's test
Bonferroni correction
之类的书,中英文的都可以。中文更好,入门嘛,便于理解。
哪位能给个提示。
谢谢 谢谢
a*****3
发帖数: 601
11
来自主题: Statistics版 - Question about Bonferroni method
赞 你e文真不是盖的
s*r
发帖数: 2757
12
来自主题: Statistics版 - Question about Bonferroni method
Pepperoni does use pork variants
A*******s
发帖数: 3942
13
来自主题: Statistics版 - Question about Bonferroni method
good one
P****l
发帖数: 156
14
来自主题: Statistics版 - Question about Bonferroni method

哈哈~~~agree
l*********s
发帖数: 5409
15
来自主题: Statistics版 - Question about Bonferroni method
zan :-)
h******3
发帖数: 190
16
来自主题: Statistics版 - Question about Bonferroni method
> <
我知道错了。大家不要围观初学者。
可能学的时候的例子是用pooled variance,所以造成错误印象.
N****n
发帖数: 1208
17
来自主题: Statistics版 - Question about Bonferroni method
全忘了。。。。要复习,好好准备面试ING 。。。
f***a
发帖数: 329
18
来自主题: Statistics版 - Question about Bonferroni method
赞,哈哈~~!
x**********0
发帖数: 163
19
I have a small sample size (N=20; 10 in intervention group, 10 in control
group) and too many variables (7 dvs), that's why I can't run a Manova.
The big overall question is always whether or not the intervention group
makes more progress and this is rarely the case.
I was advised to run separate mixes method Anovas with Time (pre-post)
and group as in between factors.
My biggest issues are the questions about Bonferroni corrections for
multiple ANOVAs- can I run the analyses by research questio... 阅读全帖
k*******a
发帖数: 772
20
用4,不过这种情况用 Tukey可能更好
d***2
发帖数: 341
21
数据有四组, 可是病人只有两组呢, 为什么需要用到multiple comparison methods?
A*******s
发帖数: 3942
22
thanks for helping me refresh my memory. i almost forget all of them...
A*******s
发帖数: 3942
23
来自主题: Statistics版 - one question about variable selection in SAS
又学习新东西了,谢谢大牛
包子sent
不过突然有个疑惑,
比如说在logistic里面
我想test H0: beta2=0, beta3=0;
这个collection是直接做一个likelihood ratio test拉倒,还是做俩wald tests with
Bonferroni adjustments ?
c*******1
发帖数: 342
24
来自主题: Statistics版 - one question about variable selection in SAS
likelihood ratio test.
Bonferroni is conservative, right?

with
A*******s
发帖数: 3942
25
来自主题: Statistics版 - one question about variable selection in SAS
yes LR is less conservative.
But LR requires refitting the model, correct?
Wald statistic with Bonferroni doesn't.
Not sure how SAS implements it.
R*****g
发帖数: 99
26
来自主题: Statistics版 - 请教个基本的统计问题
? multiple proportional tests不是可行么?
不过这里好像不需要adjust a level呀,一般来说如果需要take care multiple tests
的问题,用bonferroni

吗?
z**********i
发帖数: 12276
27
来自主题: Statistics版 - 请教个基本的统计问题
好比ANOVA要BONFERRONI ADJUSTMENT,多组比较要避免(1-0.05)(1-0.05)(1-0.05)...
不然,得到的结论太不保守.
如果,我理解错误,请指教.
l*********s
发帖数: 5409
a****g
发帖数: 8131
29
就是觉得bonferroni太conservative了
a****e
发帖数: 150
30
You can either do a Cochran-Armitage test for trend or pair-wise test with
Bonferroni correction. If only want to test independence do a chi-square
test with df=2.
s*********y
发帖数: 34
31
pair-wise test iwth Bonferroni correction应该是适用于numberical response.我
这里是Y/N的response. 不过在SAS网页看了一下Cochran-Armitage test,能测出
response是否随X(X = samll, medium, large)的增加而增加/减少. 感觉还是在测
linear trend for ordial data.
s*r
发帖数: 2757
32
来自主题: Statistics版 - 我问一个极其sb的问题。
"那我们分析大脑数据的时候,把这N个病人当成random sample;分析心脏方面的时候
,还是。"
what is wrong with this.
你又不是同时分析大脑和心脏.
it may become a problem if you want to establish certain family wise type I
error across endpoint: one obvious issue is bonferroni may be over-
correction. on the other hand, if you really want to establish a
significance between one X variable and many endpoints, you should run a
repeated measure analysis.
z******n
发帖数: 397
33
对于第一个问题,从实用的角度来讲,是不是这样描述问题更有意义?
假设存在某个CTR的值,例如5%,超过这个值的广告被业界认为是成功的,然后作如下
检验:
H0:CTR_i >= 5%
这个检验很简单,大样本时可以用正态近似,小样本时用二项检验查表,每个广告算个
p-value,然后用bonferroni correction或者其他多重检验校正,这样可以得到一批“
成功”的广告列表
对于比较多个proportion间的差异,理论上可以用Marascuillo test,但是这个检验的
零假设太无聊,更遑论chi-sq test for homogeneity of proportion了
对于这个方法选出的广告,再做两两比较(还要做多重检验校正),这时的问题是得出
的结论可能是矛盾的,比如A比B好,B比C好,但A不比C好之类的。但其实问题不大。
m*******1
发帖数: 855
34
来自主题: Statistics版 - Bonferroni correction
哈哈,总算有个认识的名次了~~~但是。。。。我不知道答案 -_-, 我只知道在
decision tree 中,如果一个变量有很多level的话,就要用这个东东
c**d
发帖数: 104
35
来自主题: Statistics版 - Bonferroni correction
It depends on your research goal.
For me, for example,
if your primary goal is to find relationship between R and Q1, Q2, ...QM,
you'd better adjust for it.
On the other hand, if it is your secondary analysis, you may not need to
adjust for it.
c****y
发帖数: 94
36
来自主题: Statistics版 - 请教一个统计问题 (转载)
It sounds like that you did a pairwise comparisons among those methods. You
need to apply adjustment. You actually should have 10 pvalue (5x4/2). If you
want to use Bonferroni Correction, it may be too conservative.In sas, you
can use proc multtest with hommel option (different way to control family
error rate). Hope it can work for you. If adjusted p-value is not
significant, it just means there are not significant difference between the
methods (2 methods compared). Actually, you should do ANO... 阅读全帖
v*******e
发帖数: 11604
37

规章制度不知道啊,科学上说,如果用Bonferroni弄过的P value是可以的。比如总共
有100种subgroup可能性,那么某一个subgroup的P value达到0.0005可以说得过去。
A**H
发帖数: 4797
38
谢谢
我根据这里的方法做的clustering
http://www.statmethods.net/advstats/cluster.html
用的其中的Partitioning这一节,我得到了一个"Within groups sum of squares"
versus "Number of Clusters" plot. 从这个plot里面我选了clusters = 5,然后做下面
fit <- kmeans(mydata, 5) # 5 cluster solution
# get cluster means
aggregate(mydata,by=list(fit$cluster),FUN=mean)
# append cluster assignment
mydata <- data.frame(mydata, fit$cluster)
我得到了哪个项目应该归到哪一类里面
感觉到这里,似乎就已经做完了。。。。。我知道了哪些项目应该编排到一起
然后,我再根据下面这个
http://www.stat.columbia.edu/~martin/W2024/R3.pdf
做anov... 阅读全帖
首页 上页 1 2 (共2页)