Help analyzing Split Plot with SubPlot in Fit Model with ANOVA for each factor.

mavargo

I have an experiment where I grew plants on three separate benches in a gh and measured the response of shoots harvested on flowering time.

My whole plot factor is light (3 levels) and my blocking factor is the benches (1-3)

On each of these benches,  each light level is present (randomized on each bench) (1-3)  So I have replication on each bench. With each light level treatment replicated 3 times.

Each row represents the benches (1-3) with each light level present on each bench.

Light level 3	Light level 2	Light level 1
Light level 2	Light level 1	Light Level 3
Light level 1	Light level 3	Light Level 1

Within each of these light levels on each of these benches, I had 36 plants in which I was applying a chemical at 4 different rates:

Split-plot factor: Chemical application rate: 4 levels.

With 36 total plants, there were 9 plants per chemical treatment in each light level.

Sub plot factor: Harvest position

Those 9 plants receiving a specific chemical treatment were then subdivided into where I was harvesting shoots.

I  have a base location on the plant called primary stalk nodes 1, 2, or 3 and harvested the actual shoots that arose from each of the primary nodes at two different secondary node positions, 2 or 3.

 

So, 3 of the 9 plants were harvested at primary nodes 1-3 at sec 2 shoots that arose from each of the primary's.

the remaining 3 plants of the 9 were also harvested on primary nodes 1-3 but I took the 3rd node shoot that arose from each of the primary nodes. So I have combinations of node positions for harvesting shoots. 

Primary	Secondary
1	2
2	2
3	2
1	3
2	3
3	3

 

I am very confused and I do not know how to set my data up in JMP. I have been reading lots of threads that say I need to nest my light by tables and it is a random effect, etc, etc. but I am still not sure what is the correct way to set it up and what attributes are associated with each factor.

Can someone explain how I set this up and which model to run?

 

cwillden

Looks like you have what I would call a "blocked split-split-plot" experiment, which is a little nuts.  If I understand your problem correctly, you would need to do the following in Fit Model:

1. Add Bench to your model as a random effect.

2. Add a "Whole Plot" block as another random effect.  This should group together data points from the same light level on the same bench.

3. Add a "Subplot" block as another random effect, which groups together data points from the same combination of bench, light level, and chemical application rates.

4. Add a "Subsubplot" block as another random effect, which should uniquely identify individual plants.

5. Add all your fixed effects (Light Level, Chemical Application Rate, and Harvest Position, any desired interaction terms, etc.).

 

If you upload your data table (delete the response data if you want), I can better help you.