Nested effect

[#1]

I have data for two species and two types of leaves. Species A and B, type R and L. I would like to check how species and type of leaves affects for example leaf area. However I have results of area measurements for several leaves assigned to each individual.

Because of that I used nest function:

Model effects:

Species

Type

Species*type

Leaf number [Individual]

But I receive “The model is missing an effect Intercept [Individual]”

What am I doing wrong?

When I mark it as random effect: Leaf number [Individual] & Random I obtain results without communicate from software, but I’m not sure if it’s entirely correct in my model.  I have a set of leaves assigned to each individual. Each leaf and individual are not connected randomly.

I just want the JMP to “see” that area results for each set of leaves should be connected with actual individual.

Do you know how to obtain that?

[#2]

Hi,

Questions about random effects seem to come up regularly in the community discussions. You are not alone!

By specifying a Random attribute for the nested effect of Leaf Number within Individual you are not saying that they are "connected randomly." What you are saying is that there is likely to be variation in leaf area from one leaf to another (for a given individual).

If you specified this as a fixed effect (the regular, default flavour of effect) you would be saying that there is a systematic difference between one leaf number and another. For example, for any individual, area is greater for leaf #2 vs leaf #1. I don't think that is the effect that you are expecting to see or that you wish to model.

By specifying the effect as Random you are acknowledging that, within an individual, there is random variation in area from one leaf number to another.

You might also want to include Invidual as a random effect if you think there could be a difference in area between individuals that needs to be accounted for.

Hope this makes sense. Explaining random effects is not simple.

Phil   

[#3]

Just another thought that might help.

If leaf number related to a position on the individual (e.g. #1 is at the base, #3 is at the top and #2 is midway) then you might expect a systematic difference in leaf area between #1, #2 and #3 that would be consistent across individuals. In this case you would want to specify Leaf Number as a fixed effect to estimate the effect of position on individuals.

If leaves are collected at random from each individual and then assigned leaf numbers you would not expect any systematic difference in area between leaf numbers. There would be differences between leaf numbers but it should be entirely random. Therefore you would specify Leaf Number as a random effect to account for this variation.