what do you mean multivariate ANOVA?



所有跟贴·加跟贴·新语丝读书论坛http://www.xys.org/cgi-bin/mainpage.pl

送交者: xj 于 2005-8-19, 23:08:27:

回答: Who is good at Multivariate ANOVA? Need help. 由 cornbug 于 2005-8-19, 22:34:59:

I like to help you but please first clarify what you mean Multivariate ANOVA.

Usually when we talk about MANOVA, we assume you have several response variables (Ys) somewhat correlated, and sure you have one or more categorical indepedent variables (Xs). The big assumption (and also advantage) of MANOVA is that Ys are multivariate normal distributed, and MANOVA will take this correlation into account.

As an superficial example, if you feed mice using several different diets, and you measure the weight and percent of fat as depedent variables, since weight and %fat are correlated, the MANOVA is better than two univariate ANOVA.

One interest in MANOVA is to compare group difference in all these response variables, that is, treat these response variables as an unseparatable set. A general F test can be obtained in most software.

You may also be interested in post hoc difference between groups for each response variable, which can be obtained easily through simple test.

Theoretically, MANOVA falls into the general linear model (GLM), and SAS can do it using proc GLM. Just put all your response variables together in one model statement and you are done.

I am an applied person, please don't ask me to prove anything, as it is impossible through internet. :-)





所有跟贴:


加跟贴

笔名: 密码(可选项): 注册笔名请按这里

标题:

内容(可选项):

URL(可选项):
URL标题(可选项):
图像(可选项):


所有跟贴·加跟贴·新语丝读书论坛http://www.xys.org/cgi-bin/mainpage.pl