i*****0
发帖数: 34 | 1 I have a question regarding statistics in experiment design. My supervisor
said that the replicates n at least equal 3 in order to publish paper. But
if CV is small enough ( < 5%), dupilicate (n=2) can be acceptable? I will
greatly appreciate it if somebody can give me an answer about this. |
a***h
发帖数: 1222 | 2 How do you get CV when you have less than 3 samples?
【在 i*****0 的大作中提到】
: I have a question regarding statistics in experiment design. My supervisor
: said that the replicates n at least equal 3 in order to publish paper. But
: if CV is small enough ( < 5%), dupilicate (n=2) can be acceptable? I will
: greatly appreciate it if somebody can give me an answer about this.
|
i*****0
发帖数: 34 | 3
CV=Standard Error/mean
standard error = standard deviation/(n)^0.5
n=2 still have standard deviation
what's wrong with this?
【在 a***h 的大作中提到】
: How do you get CV when you have less than 3 samples?
|
a*****h
发帖数: 176 | 4 CV = Cross Validation
我认为 n>=3 是让你的样本数足够多, 不是说能不能得到 standard error。 如果n=1
你也能得到 standard error.
【在 i*****0 的大作中提到】
:
: CV=Standard Error/mean
: standard error = standard deviation/(n)^0.5
: n=2 still have standard deviation
: what's wrong with this?
|
i*****0
发帖数: 34 | 5 n=1 you can't get standard error, because there is no standard deviation for
n=1
1
【在 a*****h 的大作中提到】
: CV = Cross Validation
: 我认为 n>=3 是让你的样本数足够多, 不是说能不能得到 standard error。 如果n=1
: 你也能得到 standard error.
|
k**w
发帖数: 963 | 6 STD=standard deviation
从没见人用CV的 |
a*****h
发帖数: 176 | 7 理论上应该是0吧
for
【在 i*****0 的大作中提到】
: n=1 you can't get standard error, because there is no standard deviation for
: n=1
:
: 1
|
m*l
发帖数: 507 | 8 这个东西不是个纯理论的问题，还有个习惯在里面。
你看看美国的人口普查，费了多少年的事才实现“理论”上的正确啊。
我觉得取三个样基本上也是这样，不管理论上如何，已经是大家约定俗成的东西了。做
了三四年，最后因为少取一个样被编辑给枪毕了，不划算啊。 |
