M*****c
发帖数: 3306 | 1 一个实际应用题目，比如我总共有200根元件，每个元件的长度比正常多出或者少了几
公分（所有数据都知道了），我现在从当中抽出10根（长度也都知道了）做力学实验，
做好结果以后呢，我担心这10根元件有没有代表性，请问统计上怎么计算啊？我记得有
个抽样代表性的公式，能告诉我一下吗？别给我太多数学符号，我不是学数学的。谢谢
。 |
e******o
发帖数: 644 | 2 其实就是一个从1-200不重复取10个样本 R中的程序如下
x<-c(1:200)
y<-sample(x, size=20, replace = FALSE, prob = NULL)
没有什么代表性一说 如果你认为这10根元件没有代表性 除非显然样本存在分层情况
那么就要stratified sample了
【在 M*****c 的大作中提到】
: 一个实际应用题目，比如我总共有200根元件，每个元件的长度比正常多出或者少了几
: 公分（所有数据都知道了），我现在从当中抽出10根（长度也都知道了）做力学实验，
: 做好结果以后呢，我担心这10根元件有没有代表性，请问统计上怎么计算啊？我记得有
: 个抽样代表性的公式，能告诉我一下吗？别给我太多数学符号，我不是学数学的。谢谢
: 。
|
s*r
发帖数: 2757 | 3 you can just compare means and variances
you can also test whether the empirical distributions estimated from the 200
and the 10 samples are identical |
M*****c
发帖数: 3306 | 4 Here is what I did. I guess it makes sense although I am sitll not very
clear about something.
I use significance level afa=0.05, and get k=1.96.
then I calculate
K = |x-nu|/sigma/sqrt(n)
where
x is the average of 10 samples;
nu is the average of 200 samples;
sigma is standard deviation of 10 samples;
n = 10
K = 0.46 < k, so I think it is a good representative. I think it is
basically a Hypothesis Test problem, right?
But my ques |
D******n
发帖数: 2836 | 5 no two things are the same in the world,
10 out of 100 is of course not the same as the original 100.
the problem is how close, ur afa is controlling that precision. But how
close is close is up to you to decide.
by the way as u already know the actual length of the unit, u can use it as
a co-variate to reduce the variance.
【在 M*****c 的大作中提到】
: Here is what I did. I guess it makes sense although I am sitll not very
: clear about something.
: I use significance level afa=0.05, and get k=1.96.
: then I calculate
: K = |x-nu|/sigma/sqrt(n)
: where
: x is the average of 10 samples;
: nu is the average of 200 samples;
: sigma is standard deviation of 10 samples;
: n = 10
|
M*****c
发帖数: 3306 | |
