|
Hello. I am trying to analyse a data set with
multiple dependent variables in a nested design. This requires a nested
MANOVA and I don't know how to specify the error structure. I am using
SPLUS 6. Any help is welcome.
Data structure: A field is divided into 4
blocks. Each block contains 4 parcels and one of 4 levels of a treatment
is randomly assigned to each parcel in each block. 6 soil cores are
randomly extracted from each parcel and each soil core is analysed in
triplicate. There are 26 dependent variables (abundances of each of 26
"species" of soil bacteria).
Analysis: Because of the nested structure, the
treatment effect has to compare the responses between treatments with the
responses between parcels within treatments (similarly for the block
effect). Systematic differences between soil cores within a parcel would
be tested by comparing the between-cores within parcels responses to the
between-replicates within cores within parcels responses.
Model (I think...): cbind(y1,y2 etc) ~ block +
treatment + block:treatment + cores%in%block:treatment
Problem:
How do I specify that the treatment and block
effects are to be tested using the block:treatment term as the error and not the
default residuals (which are the between-replicates within cores within
parcels)?
|