For the last 3 weeks I have been reading books, papers and internet sources incl. cross validated but could not find something that would work in my case.
I have data from a split-plot field trial. The field trail was laid out in the same way at 2 locations over 2 years. The literature said that it would be best to do the statistics in SAS with PROC MIXED because GLM would give me some false results.
So my question would be what would be the best mixed model for 2 years and 2 locations.
Fixed factors are:
- tillage as the main plot factor
- plant as the subplot factor
- location
Random factors are:
- block
- year
For one year and one location I have managed to get a model as it can be found in the "hitchhiker's guide to mixed models for randomized experiments".
The model for one year one location would look like this.
Model yield = tillage plant tillage*plant;
Random = block block*tillage block*tillage*plant
The next step would be the model for 2 years at one location, but for this case I could not find any literature that would work for a split plot. Because every year there was a new set up for the experiment I don't think that I could consider it as repeated measurements. What should the model look like?
Step 3 would be the model over 2 locations. I have read that locations can be seen as 'superblocks' but don't know how I should build the model for the 2 year 2 location split plot.
I would be so happy if somebody could help or point me in the right direction where I could find some models like this. Thanks in advance :)
