Started a mixed effects model based on split plot two factor data.

[moorepants]

@msimmond Here is a start of something for a mixed effects example. I don't really know what to do with it though.

In the design doc we listed that we want to do a mixed effect model for an RCBD design with a single measured variable. I'm not sure how you even get an RCBD design with only one factor. We then wanted to support location and/or year as random effects. But isn't the block a random effect in RCBD? Anyways, this example I've started may be useless as it is a split plot design with two factors.

I haven't been successful at finding an example data set for an RCBD with one fixed factor and one/two random factors.

[msimmond]

"I'm not sure how you even get an RCBD design with only one factor."
You're right. 'Block' is also a factor. So if there is 1 treatment in a RCBD, that's actually 2 factors. So if we want a RCBD with block and treatment as fixed (how it's been so far), and year or location as random, we will actually have 3 factors. I'd like to start with that example since it's simpler, and then do the split plot example after (if that was one of the scenarios - I forget)

"I haven't been successful at finding an example data set for an RCBD with one fixed factor and one/two random factors."

I will start looking for this. Do you want to start a pull request that I can work from?

[moorepants]

I can start a pull request, but I need some example. Should I take the RCBD dataset we already have and just fabricate some location data?

[msimmond]

Yes, sounds good.

[moorepants]

Ok, I'll push something up to start with in a bit.

[msimmond]

Actually, what I'm reading is saying block is random in a mixed effects model (in addition to year/location). http://www.unh.edu/halelab/BIOL933/lectures/lect_15_reading.pdf

[moorepants]

Yeah, that makes sense to me too. Intuitively I'd think block should be random.