Re: ANCOVA

On 30 May 2006 07:31:19 -0700, "bill" <rkilada@xxxxxxxxx> wrote:

Hello all,
To compare the slope of 2 straight lines, I found many literature using
ANCOVA (parametric) without testing the normality of the 2 samples. Is
this right? if not, what is the other option?

If you cared much about "normality", the care should be
applied to the residuals of the straight lines, and not
to the original scores.

Outliers tend to mess up the testing, and not "non-normality"
by itself.

If you Google groups on < regression normality >,
I think you will find support for paying less attention
to normality, and more attention to outliers or patterns
in residuals.

--
Rich Ulrich, wpilib@xxxxxxxx
http://www.pitt.edu/~wpilib/index.html
.

Relevant Pages

  • Re: Some opposites
    ... I think that _having_ marked terms equivalent to 'gay' predates ... telling is how 'straight' is overloaded -- it's really a catchall ... but having connotations of normality and ... The relevance here being, of course, that the informal antonym ...
    (alt.usage.english)
  • Re: ANCOVA
    ... To compare the slope of 2 straight lines, ... ANCOVA without testing the normality of the 2 samples. ... The residuals should also be homoscedastic, ... both the normality and homoscedasticity ...
    (sci.stat.consult)
  • ANCOVA
    ... To compare the slope of 2 straight lines, I found many literature using ... ANCOVA without testing the normality of the 2 samples. ...
    (sci.stat.consult)
  • Re: Unsolved problems
    ... scales, it makes comparably more sense to consider r_s a suitable ... r_s is a measure of monotonic association rather ... The presence or absence of outliers is IRRELEVANT also. ... worrying about normality of variables. ...
    (sci.stat.math)