# Nested effects in lme4

I have one park where each half is under different management (due to county lines splitting it) and I'm interested in examining the effect of the management style on individual subjects. Within each side of the park I have two clans, subjects belong to one clan.

score ~ park + (1|park/clan)

However, when I run this model the variance and SD are at 0 for park:patch in the model summary. In addition I'm getting the same results as when I just run

score ~ park

Below is a plot of my data (error bars show 95% CI). Basically I just want to know: are the two parks significantly different?

• You begin your questions with: "I have one park..." and end with "...are the two parks significantly different?". I assume the latter should read: "are the two management systems significantly different?" Is this correct? – Frans Rodenburg Oct 8 '17 at 3:39
• Yes, I realize that was confusing. In my model I was using the word "park" to refer to management system. You are correct in your interpretation. Thank you! – Lily Oct 11 '17 at 7:14

If you have one park, you cannot add it to your model. You said you have two management systems, so I interpret your problem as follows:

If you want to incorporate the effect of subjects being in clans in the park, your model would look something like this (taken from your description):

score ~ management.system + (1|clan) for a random intercept
score ~ management.system + (0 + management.system|clan) for a random slope
score ~ management.system + (management.system|clan) for both

The interest lies in the management systems and you didn't mention them being part of some larger population of management systems you wish to generalize the results to, so I see no reason to model it as a random effect.

There is no nested effect here, unless there are multiple measurements of each subject, in which case the model should be:

score ~ management.system + (1|clan/subject)

Or its equivalent for a random slope or both.

I hope this answers your question satisfactory.