Skip to content

contrast of predicted slopes from MCMCglmm model

1 message · Daniel Fulop

Original
Daniel Fulop
# 
Hi mixed modelers,

I would like to contrast average slopes using predictions from an LMM 
model fit with MCMCglmm.  In essence I want to do something analogous to 
lsmeans' lstrends() function.  The predicted slopes are growth rates 
...more below.

I have 12 days of plant growth data for 10 closely related species at 2 
temperatures.  There are several individuals per species at each 
temperature, and they're fully randomized.  I want to asses whether each 
species' growth rate differs between temperatures.  What I mean by that 
is that I would like to contrast the average slopes for each species in 
control vs. cold temperature; these slopes are the growth rates.

I am modeling the data in MCMCglmm so I can account for phylogeny and 
also to be able to try multi-response models (to jointly model the stem, 
plastochrons, root, etc).  I am starting out with the stem length data 
and after trying different time dependencies settled on this 3rd degree 
polynomial model:

stemLen ~ poly(day, 3, raw=TRUE) * treatment * species + (1 + day | indiv)

where day is a numeric time variable, treatment a 2 level factor 
(control and cold), and species a 10 level factor.

I have a slight idea of how I would go about contrasting the average 
slopes for each species and that I would use the predict.MCMCglmm() 
function, but I would really appreciate some guidance before I go down 
the wrong the wrong path.

Thanks for your help!
Dan.