Advice on repeated measures

[<Andre.Mery@aventis.com>]

Dear all,

 

I would appreciate an advice on how to perform a repeated measurement analysis with "lme".

 

Basically, the data I have are as follows:

 

Animal    Housing    Group    Day    Score

(...)     (...)      (...)    (...)  (...)

 

- Housing is a factor with 2 levels

- Group is a factor with 4 levels

- Day is a factor with 12 levels

 

For each Animal (belonging to a particular Housing level and a particular Group level, 12 measures or Score are made on the 12 differents days. Data are balanced (no Housing level, no Group level, no Day level, are missing for any Animal)

 

In the treatment of similar data (the so-called "Sitka" example), Venables & Ripley (their book, p.206 in the 3rd edition) use the code:

 

lme(size ~ treat*ordered(Time), random = ~ 1 | tree, data = "">

*

*    *

1°) In my case, I want to differentiate the 2 Housing levels (is there a significant difference between them?), taking Group as a covariate. So, is it sounded to copy the previous code and to adapt it by using::

 

lme(Score ~ Group + Housing*ordered(Day), random = ~ 1 - Animal, data = "" ???

*

*    *

2°) If I do this, a call to "summary" gives me:

 

Fixed effects: Score ~ Group + Housing * ordered(Day)
                           Value Std.Error  DF   t-value p-value
           (Intercept)  1.512500 0.0881239 858  17.16333  <.0001
                 Group -0.021667 0.0203514  77  -1.06463  0.2904
               Housing -0.016667 0.0455070  77  -0.36624  0.7152
        ordered(Day)^1 -0.386762 0.1626184 858  -2.37834  0.0176

        (...)

Housingordered(Day)^11  0.082391 0.1028489 858   0.80109  0.4233

 

Can I readily conclude that Housing#1 is not different from Housing#2 (due to p-value = 0.7152) ???

*

*    *

 

I am not really sure of what is the proper treatment of repeated data with "lme". Taking the format of my data, I would appreciate to know whether you think a simple treatment like the previous one is enough to get a conclusion on Housing#1 versus Housing#2, or if I should use something else more appropriate.

 

Thanks.

 

André Méry
Aventis Pharma
20, avenue Raymond Aron - 92160 Antony
[Courrier / Mail : Tri B2/13]
[Tél. / Tel. : 01 55 71 68 69]