Re: Covariate in ANCOVA question
- From: David Winsemius <doe_snot@xxxxxxxxxxx>
- Date: Tue, 23 Oct 2007 22:35:43 -0500
andre_xs@xxxxxx wrote in news:1193069172.326440.137880
@k35g2000prh.googlegroups.com:
We have sound recordings of 3 groups of bats, and our aim is to test
whether the pitch of the bat calls differs between groups. However,
the calls are a frequency sweep from a constant high pitch to a lower
pitch, so that calls with a longer duration will invariably have a
lower pitch in the end, as well as averaged across the whole call.
Since we need to analyse the average pitch of the call, we need to
account for call duration. Importantly, the groups differ
significantly in their call duration. From plots we see that they
differ in pitch as well, but how to test for that statistically?
Our idea is to use an ANCOVA with call duration as the covariate
(AFAIK in SPSS Analyze->GLM->Univariate->"Covariate"-Box). We probably
would start with an ANCOVA of all 3 groups of bats, and if there is a
statistical outcome, to use 3 individual ANCOVAs (with only two groups
of bats included) for the pairwise comparisons of all 3 groups. We
would use ANCOVAs instead of t-tests to have the option of including
the duration covariate. In addition, we would start with a global
ANCOVA including all 3 groups, since there are further parameters
besides pitch to test.
Is that a valid approach? If not, what would be advisable to do?
ANCOVA packages should also let you specify appropriate post-hoc pairwise
comaprisons. The first step would be fine but your second would be
incorrect. You would not get the desired type I error levels in the
pairwise comparisons. I suggest you search with these terms:
SPSS "analysis of covariance" "pairwise comparisons"
--
David Winsemius
.
- Follow-Ups:
- Re: Covariate in ANCOVA question
- From: andre_xs
- Re: Covariate in ANCOVA question
- From: Bruce Weaver
- Re: Covariate in ANCOVA question
- References:
- Covariate in ANCOVA question
- From: andre_xs
- Covariate in ANCOVA question
- Prev by Date: Re: Fitting Assymetric Curves?
- Next by Date: Re: Covariate in ANCOVA question
- Previous by thread: Re: Covariate in ANCOVA question
- Next by thread: Re: Covariate in ANCOVA question
- Index(es):
Relevant Pages
|