I have biomass data for different sediment depths (0-1,1-2,2-3 and 3-4 cm) from the same sediment core. At different time-steps (T0, T1, T7 and T14), one core was analysed (so no replicates). Moreover, there are no data for 3-4 cm at T14 (ie. missing data or unbalanced design). I'm interested in finding out significant differences in biomass between sediment depth layers.
I thought of using the following design
Time is the blocking factor and sediment depth is the treatment
I thought of using biomass~sediment depth + time, but this only works if there is no significant blockxtreatment interaction (which can be tested with the Tukey test for non-additivity)
My question now is
- what if there is a significant interaction? how to proceed then? I didn't find this anywhere
- shouldn't I use Anova type III sums of squares for unbalanced data?
Thank you very much!
Advertise on Talk Stats