2
$\begingroup$

I am attempting to fit a generalized linear mixed effects model in R using lme4, but am new to mixed models. I'd like to model Survivorship as the outcome of Treatment + Environmental Factors (clarified below), with random effects for Plots nested within Sites. My question and experimental design are fairly similar to this post: https://stats.stackexchange.com/questions/79360/mixed-effects-model-with-nesting.

I have data collected from an experiment organized as follows:

Two sites, each with 30 vegetation plots on a series of created berms (7 at one site and 13 at the other). In each plot, 3 seedlings of 2 different species are sampled at 2 treatment levels (T vs C) (6 of each species per plot). At each plot, environmental variables- soil moisture (SM) and elevation- are measured. The oddball, soil bulk density (BD), however, is measured at the mound level. Total number of observations is 720; 2 sites * 30 plots * (three CALA seedlings + three CAST seedlings) * (treatment 2 + treatment 3). Data look like this (not all provided)...

ID Site Plot Mound Treatment Species Survivorship SMAvg Elevation    BD
1   BC 1Abc   1bc         2    CALA            1     1   75.16    0.111
2   BC 1Abc   1bc         2    CALA            1     1   75.16    0.111
3   BC 1Abc   1bc         2    CALA            1     1   75.16    0.111
4   BC 1Abc   1bc         2    CAST            1     1   75.16    0.111
5   BC 1Abc   1bc         2    CAST            1     1   75.16    0.111
6   BC 1Abc   1bc         2    CAST            1     1   75.16    0.111
7   BC 1Abc   1bc         3    CALA            1     1   75.16    0.111
8   BC 1Abc   1bc         3    CALA            1     1   75.16    0.111
9   BC 1Abc   1bc         3    CALA            1     1   75.16    0.111
10  BC 1Abc   1bc         3    CAST            0     1   75.16    0.111

I am performing this using glmer, as follows:

M1 <- glmer(Survivorship ~ Treatment + SMAvg + Elevation + BD +
 (1|Site/Plot). 

Q1) I have not included Site as a fixed effect (although it is indeed fixed) because it has been described elsewhere that a categorical variable should not be entered if it is already included as a "random" effect. Is this accurate?

Q2) Lastly, because I have measured certain environmental variables at different levels (e.g., SM at plot level and BD at mound level), does this inherently preclude the model clarified above? Must I drop BD, therefore?

Q3) I have found Mound to be a confounding effect overall (plots are not equally dsitributed across them and there are a nonequivalent number at each site), so do not know how to deal with it. Leave out?

I am very driven to become fluent in GLMM, but could use some feedback! Thank you.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.