Skip to content
Prev 16298 / 20628 Next

Estimates for groups within fixed effects & also: enough info for estimating variance? MCMCglmm

Hi Jarrod,

Thanks so much for your reply & patience with my lack of stats/math saavy.

If fitting the model to include us(Znnd):species, I'm unsure of how to  
specify the prior. Based on your Course Notes, I've tried:
priorI= list(R = list(V = 1, nu = 0, fix = 1), G = list(G1 = list(V =  
diag(2) * 0.02, nu = 2),
+                  G2 = list(V = 1,nu = 0.002)))


Where (G1 = list(V = diag(2) * 0.02, nu = 2) represents a 2 X 2 matrix  
of distances between individuals X species Is this the correct matrix  
size, since Znnd is continuous? It was hard to extrapolate from the  
us(1+time:chick) data as there were only two time points per  
chick...but I understand that with categorical predictors, you would  
have a matrix that is the size of the number of categories.

Would it be possible to walk me through how to specify nu? For  
example, the difference between specifying nu=dim(V)+1 vs nu=dim(V)? I  
see us() examples with both and am unclear as to their relative merits  
other than their both being attempting  to be uninformative?

Also, my response variable contains high replication (>1000) for the  
majority of species but very low replication for a few (16, 20, 24),  
and is very skewed towards having more zeroes than ones overall.

I have yet to follow up on your suggestions for part 2.

Many thanks,

Andrea



Quoting Jarrod Hadfield <j.hadfield at ed.ac.uk>:
Andrea Pilar Drager
PhD. student
Ecology and Evolutionary Biology, Rice University