I'm running this in R. I have a dataset with some nested data. I have two land ownership categories and five road types nested within these landownerships. I will include a flow chart of my IVs so they can be understood better. My sample size is not massive, n = 202. However, there are repeated measures within my samples, also there are only one transect each for the private land road types. Thus, I believe I need to run a mixed-effects model. This is where I get lost, if there is nesting, how can I account for this within my model, and which IV should be random (ownership or road type)? I'm thinking that ownership should be fixed and road type should be random or should transect be random? However, when I run the model with transect as a random effect is has a variance of 0 and a SD of 0. Of course, I have other variables too but these are all fixed. Can I get some help guys?
$\begingroup$
$\endgroup$
5
-
$\begingroup$ What is a transect? $\endgroup$– Jeremy MilesCommented Mar 13, 2021 at 5:39
-
$\begingroup$ Meaning we surveyed two different transects for O, C, and WILD each, equaling 6 transects and then only 1 transect for HU roads and 1 transect for MU roads. 5 road types with only 8 tramsects $\endgroup$– Cr KCommented Mar 13, 2021 at 18:27
-
$\begingroup$ My fixed effects will be season, distance, density, and season:distance. I'm confused on what I should include as a random effect... road type, transect, or ownership, or none of them? If road type is nested in ownership should the model be fgc ~ (1 | ownership/road type) + season + distance + density + season:distance? $\endgroup$– Cr KCommented Mar 13, 2021 at 18:48
-
$\begingroup$ I don't know what a transect is still, but it doesn't sound like anything is random. $\endgroup$– Jeremy MilesCommented Mar 14, 2021 at 4:10
-
$\begingroup$ Jeremy, I should rephrase my question. My apologies. So we surveyed 6 transects consisting of 2 open roads, 2 closed roads, and 2 transects with no roads. Also, 1 transect each in areas of high use conservation roads and minimal use. On these transects, we collected bear scats to measure stress hormones. Undoubtedly there was some temporal and spatial pseudoreplication. To account for this pseduoreplication, I figured I would need to add a random effect to my model, am I incorrect in thinking this? if not, where would I include that effect. $\endgroup$– Cr KCommented Mar 15, 2021 at 15:55
Add a comment
|
1 Answer
$\begingroup$
$\endgroup$
1
From the description, there does not appear to be any need to fit random effects here. All the variables listed should be fixed effects in the model:
fgc ~ ownership + roadtype + season + distance + density + season:distance
answered Mar 15, 2021 at 8:56
-
$\begingroup$ See the response above to Jeremy, my thought was because of pseudoreplication that I would need to include a random effect to account for that. The comment above to Jeremy explains what I'm doing a little better. $\endgroup$– Cr KCommented Mar 15, 2021 at 16:25