Google Groups no longer supports new Usenet posts or subscriptions. Historical content remains viewable.
Dismiss

why need to assume the constancy of variance for regression analysis and ANOVA?

55 views
Skip to first unread message

zrl

unread,
Aug 20, 2009, 5:03:12 PM8/20/09
to
thanks.

Xt

unread,
Sep 25, 2009, 6:10:42 PM9/25/09
to
On Aug 21, 9:03 am, zrl <zrl1...@gmail.com> wrote:
> thanks.

Because the p values returned in the output, on which your conclusions
are based, are calculated from theory (F distribution) which is only
true if the variance is constant. If the variance is not constant,
the p values cannot be trusted so neither can your conclusions.

One exception is the unequal variance t test which uses a mathematical
trick to adjust the p value to its approximate true value.

0 new messages