Repeated measures

In the two solutions for the repeated measures problem given in the
original reply below, the F and p values given by aov() with the error
strata defined by Error() are different from those given by lme().
However, when one does the problem "by hand" using the standard split
plot model, the results agree with those of nlme(). The difference
between the two aov() solutions is in the partitioning of sums of
squares. Is there a ready explanation for this discrepancy?

Thanks,
Richard Plant

tolerance <- tolerance <-

+
read.table("http://www.ats.ucla.edu/stat/Splus/examples/alda/tolerance1.
txt",
+             sep=",", header=TRUE)

tolerance.long <- reshape(tolerance,

+                           varying = list(c("tol11","tol12","tol13",
+                                            "tol14", "tol15")),
+                           v.names = c("tol"), timevar = "time",
+                           times = 11:15, direction = "long")

tolerance.aov2 <- aov(tol ~ factor(male) + factor(male):factor(id) +

factor(time) + factor(time):male, data = tolerance.long)

tolerance.sum <- summary(tolerance.aov2)
tolerance.sum

                        Df Sum Sq Mean Sq F value    Pr(>F)    
factor(male)             1 0.3599  0.3599  2.6077  0.111967    
factor(time)             4 2.8326  0.7081  5.1309  0.001358 ** 
factor(male):factor(id) 14 8.2990  0.5928  4.2951 4.295e-05 ***
factor(time):male        4 0.1869  0.0467  0.3386  0.850786    
Residuals               56 7.7289  0.1380                      
---
Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1 

tolerance.list <- tolerance.sum[[1]]
tolerance.mat <- as.matrix(tolerance.list[3])
tolerance.F.male <- tolerance.mat[1,1]/tolerance.mat[3,1]
tolerance.F.male

[1] 0.607137

tolerance.df <- as.matrix(tolerance.list[1])
tolerance.p.male <- 1 -

pf(tolerance.F.male,tolerance.df[1,1],tolerance.df[3,1])

tolerance.p.male

[1] 0.4488394

Message: 68
Date: Wed, 17 Jan 2007 05:45:01 -0500
From: Chuck Cleland <ccleland at optonline.net>
Subject: Re: [R] Repeated measures
To: Tom Backer Johnsen <backer at psych.uib.no>
Cc: r-help at stat.math.ethz.ch
Message-ID: <45ADFE2D.2060208 at optonline.net>
Content-Type: text/plain; charset=ISO-8859-1

Tom Backer Johnsen wrote:

I am having a hard time understanding how to perform a "repeated
measures" type of ANOVA with R.  When reading the document found

here:

http://cran.r-project.org/doc/contrib/Lemon-kickstart/kr_repms.html

I find that there is a reference to a function make.rm () that is
supposed to rearrange a "one row per person" type of frame to a "one
row per observation" type of frame.  But that function does not seem
to be there.  Nor does the help.search suggest anything.  Is that
function buried in some package?

  I'm not able to find that function.  Perhaps that document is out of
date.

Is there  some simple documentation that might be useful somewhere?
Starting with a really simple problem (one group, two observations)?

  Here is an example showing the use of reshape() and analysis via

aov()

and lme() in the nlme package.

tolerance <-

read.table("http://www.ats.ucla.edu/stat/Splus/examples/alda/tolerance1.
tx

t",
            sep=",", header=TRUE)

tolerance.long <- reshape(tolerance,
                          varying = list(c("tol11","tol12","tol13",
                                           "tol14", "tol15")),
                          v.names = c("tol"), timevar = "time",
                          times = 11:15, direction = "long")

tolerance.aov <- aov(tol ~ as.factor(time) * male + Error(id),
                     data = tolerance.long)

summary(tolerance.aov)

Error: id
     Df   Sum Sq  Mean Sq
male  1 0.085168 0.085168

Error: Within
                     Df  Sum Sq Mean Sq F value  Pr(>F)
as.factor(time)       4  2.8326  0.7081  3.0538 0.02236 *
male                  1  0.3024  0.3024  1.3039 0.25745
as.factor(time):male  4  0.1869  0.0467  0.2015 0.93670
Residuals            69 16.0002  0.2319
---
Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

library(nlme)

tolerance.lme <- lme(tol ~ as.factor(time) * male, random = ~ 1 | id,
                     data = tolerance.long)

anova(tolerance.lme)
                     numDF denDF  F-value p-value
(Intercept)              1    56 353.9049  <.0001
as.factor(time)          4    56   5.1309  0.0014
male                     1    14   0.6071  0.4488
as.factor(time):male     4    56   0.3386  0.8508

  RSiteSearch("repeated measures") points to other examples,

functions,

and documentation.

Tom

guide.html

--
Chuck Cleland, Ph.D.
NDRI, Inc.
71 West 23rd Street, 8th floor
New York, NY 10010
tel: (212) 845-4495 (Tu, Th)
tel: (732) 512-0171 (M, W, F)
fax: (917) 438-0894



------------------------------