Hopefully someone will find this far more straightforward than me. I've been struggling to work out the correct model to use with my data for 2 16 hour days and have got as far as I can. I am analysing data from an experiment where I made measurements on leaves taken from trees that had been subject to a number of treatments. The treatments are:
CO2 - 2 levels randomly allocated within block
Teees planted in mixture or monoculture - 2 levels, sub-plots within each CO2 replicate
Species - 3 levels within mixture or monoculture
I removed leaves from 5 different places in the canopy of each of 3 trees in each Block/CO2/MixMon/Species treatment
As far as I can tell it is a blocked spli- split-plot design with leaves nested within trees and trees nested within species.
I want to perform 2 separate analyses:
1. Main effects and interactions between CO2, Mix/Mon and species on measurements made for each of the leaves.
2. Differences between leaves at different positions in the canopy and the impact of effects in 1 on this.
So far I have the following model for 1, but it doesn't seem to give the correct error term for any of the interactions with 'species':
TreeSPSS identifies individual trees within each species (i.e. 1, 2, 3), Ring_for_analysis identifies each plot for the CO2 treatment (MixMon and Species are sub and sub- sub- plots within each ring)
UNIANOVA N BY CO2 Species MixMon Block TreeSPSS Ring_for_analysis
/RANDOM=Block TreeSPSS Ring_for_analysis
/EMMEANS=TABLES(CO2) COMPARE ADJ(LSD)
/EMMEANS=TABLES(Species) COMPARE ADJ(LSD)
/EMMEANS=TABLES(MixMon) COMPARE ADJ(LSD)
/DESIGN=Block CO2(Block) Ring_for_analysis(CO2(Block)) Species TreeSPSS(Species) MixMon MixMon*Ring_for_analysis(CO2(Block)) CO2*MixMon CO2*Species MixMon*Species CO2*MixMon*Species.
Hopefully someone will be able to tell me where I'm going wrong here and what I need to do to include 'leaf position' in this model.