如何用sas做两个变量是否相等的test?

是 否 +2 论坛币

k人 参与回答

经管之家送您一份

应届毕业生专属福利!

求职就业群

赵安豆老师微信：zhaoandou666

经管之家联合CDA

送您一个全额奖学金名额~ !

立即领取

感谢您参与论坛问题回答

经管之家送您两个论坛币！

+2 论坛币

有两个变量，一共有200个obversation，
想test下这两个变量的值是否相等。

用什么方法呢？

不过这个好像不仅仅是sas code的问题了，还涉及到统计量的选择.

Wilcoxon test and t test? How do we do it in these two ways?

扫码加我 拉你入群

请注明：姓名-公司-职位

以便审核进群资格，未注明则拒绝

分享0 收藏0 回帖

关键词：test Est 如何用 wilcoxon wilcox 如何 统计

一眼瞬间 发表于 2010-3-22 02:34
有两个变量，一共有200个obversation，
想test下这两个变量的值是否相等。

用什么方法呢？

不过这个好像不仅仅是sas code的问题了，还涉及到统计量的选择.

Wilcoxon test and t test? How do we do it in these two ways?

Test what? mean, variance, distribution?

2# bobguy

Bobguy, 你又给我补课了：）

There're two ways to calculate the value of one variable. I want to know whether these two ways can genereate similar results.
Should I test all?

Mean, median, distribution, cv, variance?

How can I test those statistics?

一眼瞬间 发表于 2010-3-22 03:36
2# bobguy

Bobguy, 你又给我补课了：）

There're two ways to calculate the value of one variable. I want to know whether these two ways can genereate similar results.
Should I test all?

Mean, median, distribution, cv, variance?

How can I test those statistics?

could you clearly define your situation? the questions you have asked is very ambiguous.

first you need to decide what your inference is;
then you need to select what test(s) and what statistic(s) that is needed to conduct your inference.

for example, if you wanna test on the location parameters of two population, you can use mean or median. then t-test or wilcoxon should be used.

Isscaliu 发表于 2010-3-22 11:11

一眼瞬间 发表于 2010-3-22 03:36
2# bobguy

Bobguy, 你又给我补课了：）

There're two ways to calculate the value of one variable. I want to know whether these two ways can genereate similar results.
Should I test all?

Mean, median, distribution, cv, variance?

How can I test those statistics?

could you clearly define your situation? the questions you have asked is very ambiguous.

first you need to decide what your inference is;
then you need to select what test(s) and what statistic(s) that is needed to conduct your inference.

for example, if you wanna test on the location parameters of two population, you can use mean or median. then t-test or wilcoxon should be used.

Sorry for the ambiguity of my problem.
Yes, I tested the location parameters and I used t and wilcoxon.

It's my of my homework problem.

那你就要回去好好读一下你的textbook了。

generally speaking， t是parametric， wilcoxon是non-parametric。
要用的话两个都可以，但是t的power比wilcoxon大（一般来讲parametric比non-para要强）。
但是，最重要的是，你要弄清楚under what circumstance用哪个。
一般来讲，indepedency是必须的，equality  of variance不一定需要（t可以有unequal-variance，但是wilcoxon应该没有，忘记了。但是non-parametric应该是没有）。最最重要的，normaility。如果data是normal或者approxmately normal，就可以用t。wilcoxon是non-parametric，一般来讲non-parametric是distribution-free，就是说对distribution没有要求，可以是normal或者类似normal的symmetric distribution，也可以是其他的asymmetric（我对distribution-free的理解是这样，如有吴请大家大方指出）。
所以，先要跑proc univariate 去看看data是否服从normal distribution。然后决定用t还是wilcoxon。
希望这样讲能够解决你的疑惑！

晕，怎么全是英语啊，俺英语差，看不懂。
如果是分类资料用配对卡方；
是计量资料配对t，看均数是否为0。
200对数据，对正态分布应该是稳健的，所以分布类型就不要考虑。

对，如果是200的数据，不用考虑分布问题i了