t*d 发帖数: 1290 | 1 经常有象我这样懂点统计又不是很懂的同志在用 t-test 的时候有些不放心。看了下面
这片文章后,我放心了。
Mihai Valcu & Cristina-Maria Valcu. 2010. Data transformation practices in
biomedical sciences. Nature Methods 8: 104–105.
文章写得很烂。但是对 t-test 的介绍还凑合。下面是我的一点理解。(英文是从文章
里的原文,中文是我的理解。)
Applying a t-test is now so routine that many biologists may have forgotten
that data should meet certain assumptions, and reminders of its correct use
have been published. 质疑 t-test 可能被烂用了。
For example, for a valid two-sample t-test, the assumptions are that the
samples are independent and drawn from populations with equal variances, and
that the variable is normally distributed in each group. 问题在于大家伙对用
t-test 的前提不太清楚:独立采样,方差相同和正态分布。不过这些真的是大问题
吗?
Although the robustness of Student's t-test to the violation of these
assumptions is a matter of debate, they are seldom verified and data are
sometimes transformed in ways that guarantee that the assumptions are no
longer met. 原来非正态分布的样本也可能用 t-test,尽管有争论。
The use of 'robust' versions of the t-test (Welch t-test) that do not
require the assumption of equal variance. 大家要记得用 Welch t-test,这样方
差相同这个前提也不需要了。
所以对俺们生物 wsn 来说,只要记得独立采样(这个比较大部分时候是可以做到的)
,就可以放心用 t-test 了。兔年快乐!:) | n***w 发帖数: 2405 | 2 if variances are not the same, t-test still can be applied by adjusting the
equation. I do not the name is "Welch" t-test.
In most cases, I think the sample size matters which make us turn to other
methods. Also paired t-test are pretty common. | s******s 发帖数: 13035 | 3 well, independent, normality, equal variance
1. independent这个没辙,必须遵守
2. normality, 基本上t-test很robust,因为实际比较的
是sample mean, 这个归于central limit thereom, 就是
数据够多,只要不是一些特别夸张的分布,都可以当normal.
基本上n<15要比较注意normality,或者转成non-parametric;
n>50基本上用t-test没问题.
3. equal variances. 这个stat101都应该说过,除非有理论
支持,否则都要做F test验证;F不通过的话,要用unpooled
t-test, 估计就是你说的welch t
forgotten
use
【在 t*d 的大作中提到】 : 经常有象我这样懂点统计又不是很懂的同志在用 t-test 的时候有些不放心。看了下面 : 这片文章后,我放心了。 : Mihai Valcu & Cristina-Maria Valcu. 2010. Data transformation practices in : biomedical sciences. Nature Methods 8: 104–105. : 文章写得很烂。但是对 t-test 的介绍还凑合。下面是我的一点理解。(英文是从文章 : 里的原文,中文是我的理解。) : Applying a t-test is now so routine that many biologists may have forgotten : that data should meet certain assumptions, and reminders of its correct use : have been published. 质疑 t-test 可能被烂用了。 : For example, for a valid two-sample t-test, the assumptions are that the
| a****a 发帖数: 3411 | 4 QQ plot ---normality.我觉得正态分布的检测很麻烦,很多实验数据根本不够检测正态
性。用k-s test 估计大部分都是非正态。数据少了如果你测出来非正态,难道就是真的
吗?
levene's test ---eqality of variance
以前做过某细菌的双色荧光体系,每次在荧光显微镜下看到细菌的荧光颜色比例不均一
。但是比较对照组已经有明显差异。那时候的我的导师硬说我没有达到全部都是同色,
实验不理想。现在想想,这样的实验怎么能够做到。除非作弊。很多搞生物的没有概率
和统计的思维方式,可惜很多这样的人都是pi
y
经常有象我这样懂点统计又不是很懂的同志在用 t-test 的时候有些不放心。看了下面
这片文章后,我放心了。
Mihai Valcu & Cristina-Maria Valcu. 2010. Data transformation practices in
biomedical sciences. Nature Methods 8: 104–105.
文章写得很烂。但是对 t-test 的介绍还凑合。下面是我的一点理解。(英文是从文章
里的原文,中文是我的理解。)
Applying a t-test is now so routine that many biologists may have forgotten
that data should meet certain assumptions, and reminders of its correct use
have been published. 质疑 t-test 可能被烂用了。
For example, for a valid two-sample t-test, the assumptions are that the
samples are independent and drawn from populations with equal variances, and
that the variable is normally distributed in each group. 问题在于大家伙对用
t-test 的前提不太清楚:独立采样,方差相同和正态分布。不过这些真的是大问题
吗?
Although the robustness of Student's t-test to the violation of these
assumptions is a matter of debate, they are seldom verified and data are
sometimes transformed in ways that guarantee that the assumptions are no
longer met. 原来非正态分布的样本也可能用 t-test,尽管有争论。
The use of 'robust' versions of the t-test (Welch t-test) that do not
require the assumption of equal variance. 大家要记得用 Welch t-test,这样方
差相同这个前提也不需要了。
所以对俺们生物 wsn 来说,只要记得独立采样(这个比较大部分时候是可以做到的)
,就可以放心用 t-test 了。兔年快乐!:)
【在 t*d 的大作中提到】 : 经常有象我这样懂点统计又不是很懂的同志在用 t-test 的时候有些不放心。看了下面 : 这片文章后,我放心了。 : Mihai Valcu & Cristina-Maria Valcu. 2010. Data transformation practices in : biomedical sciences. Nature Methods 8: 104–105. : 文章写得很烂。但是对 t-test 的介绍还凑合。下面是我的一点理解。(英文是从文章 : 里的原文,中文是我的理解。) : Applying a t-test is now so routine that many biologists may have forgotten : that data should meet certain assumptions, and reminders of its correct use : have been published. 质疑 t-test 可能被烂用了。 : For example, for a valid two-sample t-test, the assumptions are that the
|
|