R help - May 2003 - Summary statistics & plots of repeated measures data

I'm an R novice and my colleagues are about to convince me to get my
data into SPSS, which will presumably be easier for someone who
doesn't live in R to point and click his way into some kind of
analysis that might be meaningful.

I've got two groups of subjects (classkey in the table below).
They've each received several different treatments.  One measure is a
1-7 rating taken several times per treatment (about 1-14 times per
session).   studentkey, classkey, and treatment are factor()s.

The table looks something like this:

   ete classkey studentkey treatment
1    7        4        108       bp1
2    4        4        117       bp1
3    6        4        120       bp1
4    6        4        105       bp1
5    3        4        100       bp1
6    3        4        100       bp1
7    4        4        107       bp1
8    3        4        100       bp1
9    7        4        107       bp1
10   4        4        107       bp1

I'd like to see the effects of each of the treatments for this
within-subject comparison.  Repeated measures ANOVA seems like the
analysis I need.  The results of

summary(lme(ete ~ treatment, data=allitems, random=~1 | studentkey,
          subset=allitems$classkey==4))

follow, but I'm not quite sure what to make of them.  In particular,
I'm very confused about the meanings of the numbers in the Value
column, as they bear no relation to the group means of the data in
each of those treatments.

Linear mixed-effects model fit by REML
 Data: allitems 
  Subset: allitems$classkey == 4 
   AIC  BIC logLik
  2035 2065  -1011

Random effects:
 Formula: ~1 | studentkey
        (Intercept) Residual
StdDev:        1.39     1.58

Fixed effects: ete ~ treatment 
                Value Std.Error  DF t-value p-value
(Intercept)      5.44     0.322 493   16.90  <.0001
treatmentbp2    -0.80     0.204 493   -3.95   1e-04
treatmentbprog1 -1.84     0.214 493   -8.61  <.0001
treatmentbs1    -2.17     0.291 493   -7.44  <.0001
treatmentbs2    -1.31     0.221 493   -5.91  <.0001
 Correlation: 
                (Intr) trtmntbp2 trtmntbp1 trtmntbs1
treatmentbp2    -0.344                              
treatmentbprog1 -0.331  0.503                       
treatmentbs1    -0.239  0.385     0.342             
treatmentbs2    -0.327  0.514     0.467     0.352   

Standardized Within-Group Residuals:
   Min     Q1    Med     Q3    Max 
-2.888 -0.666  0.102  0.722  2.341 

Number of Observations: 521
Number of Groups: 24 
       
I'm clearly misunderstanding something.  This is very likely the type
of analysis I'll be doing for much of my career, I'd love to figure
out how to do it in R now.  (I've got  MASS3, & Dalgaard's Intro
Stats
with R as well as various online documents.  Pointers to relevant
sections therein would also be appreciated.)

Thanks.

-- 
Jay Pfaffman                           pfaffman at relaxpc.com
+1-415-821-7507 (H)                    +1-415-812-5047 (M)

Jay Pfaffman <pfaffman at relaxpc.com> writes:
> I'd like to see the effects of each of the treatments for this
> within-subject comparison.  Repeated measures ANOVA seems like the
> analysis I need.  The results of
> 
> summary(lme(ete ~ treatment, data=allitems, random=~1 | studentkey,
>           subset=allitems$classkey==4))
> 
> follow, but I'm not quite sure what to make of them.  In particular,
> I'm very confused about the meanings of the numbers in the Value
> column, as they bear no relation to the group means of the data in
> each of those treatments.
They're not supposed to. Not in an unbalanced design. Mostly, if there
is a substantial variation between students, the results would be
closer to that obtained from an additive linear model treating student
effects as fixed (lm(ete ~ treatment+factor(studentkey), ...)). The
treatment means is what you get if you assume zero student variation,
and in general you get some sort of average between the two.

You'll bump into those issues whatever program you choose...

-- 
   O__  ---- Peter Dalgaard             Blegdamsvej 3  
  c/ /'_ --- Dept. of Biostatistics     2200 Cph. N   
 (*) \(*) -- University of Copenhagen   Denmark      Ph: (+45) 35327918
~~~~~~~~~~ - (p.dalgaard at biostat.ku.dk)             FAX: (+45) 35327907