s-news
[Top] [All Lists]

MANOVA: specifying contrasts properly

To: s-news@lists.biostat.wustl.edu
Subject: MANOVA: specifying contrasts properly
From: Hillary Robison <hrobison@unr.nevada.edu>
Date: Tue, 4 Sep 2007 22:02:34 -0700
User-agent: Internet Messaging Program (IMP) 3.2.8
Hi all,

I sent a post out last week, but I forgot to leave my e-mail address
(hrobison@unr.nevada.edu).  In case any of you have insights into my question,
I am re-sending my post.

I having trouble specifying MANOVA (in the GUI of S-plus 6.2) that allows the
contrasts I'd like.

I was hoping the MANOVA would let me test for differences among levels of "Pop"
(my X variable with 32 levels) with respect to my repeated measurements for
each "Pop", my Ys ("logk2.k1", "logk3.k1", "logk4.k1", "logk5.k1", "logk6.k1",
"logk7.k1", "logk8.k1", "logk9.k1", "logk10.k1", "logk11.k1").  My Ys were
measured 5 times in each "Pop."  The output gives me an overall p-value for
"Pop", but does not give me p-values for the contrasts among my levels of
"Pop."

How do I specify this?

What I did in the S-plus GUI was:

Statisitics -> Multivariate -> MANOVA dialog box

In the "Model" tab of the MANOVA dialog box I specified:
Dataset: group
Formula : cbind(logk2.k1, logk3.k1, logk4.k1, logk5.k1, logk6.k1, logk7.k1,
logk8.k1, logk9.k1, logk10.k1, logk11.k1) ~ Pop

In the "Options" tab of the MANOVA dialog box I specified:
Assign Contrast: Helmert
to Variables: Pop
Contrasts: Pop=contr.helmert

In the "Results" tab of the MANOVA dialog box I specified:
Short Output
ANOVA Table
Testing with "pillai"
Estimated Coefficients
Estimated K Coef. for K-level Factor
Save In: group
Fitted Values
Residuals

The Report gives me:

Call:  manova(formula = cbind(logk2.k1, logk3.k1, logk4.k1, logk5.k1, logk6.k1,
logk7.k1, logk8.k1, logk9.k1, logk10.k1, logk11.k1) ~ Pop, data = clusters,
contrasts = list(Pop = contr.helmert), na.action = na.exclude)

Terms:
                Pop Residuals
Deg. of Freedom  32       132

Estimated effects are balanced

Analysis of Variance Table:
           Df Pillai Trace approx. F num df    den df    P-value
      Pop 32  4.426        3.275       320      1320         0
Residuals 132

Followed by the estimated coefficients for "logk2.k1", "logk3.k1", "logk4.k1",
"logk5.k1", "logk6.k1", "logk7.k1", "logk8.k1", "logk9.k1", "logk10.k1",
"logk11.k1" for each "Pop."

Thank you for your thoughts on this.

Hillary Robison (hrobison@unr.nevada.edu)

<Prev in Thread] Current Thread [Next in Thread>
  • MANOVA: specifying contrasts properly, Hillary Robison <=