[R-sig-ME] HPDinterval - Model simplification

[Sam]

Dear List,

I am experimenting using mcmcGLMM and have a question -

I am sorry i am not very statistically minded! 

So i have run a model 

MCMCglmm(group~1+ 2 + 3+ 4+ 5 + 6 + 7 + 8 + 9, random=~ a+a:b, family="ordinal", prior=prior, data=group)

1. Group has 5 levels and what i am trying to do is ask - if you are in group 1 (for instance) which factors are most significant (1,2,3,4,5 etc)

2. I then want to reverse this and ask  - if an individual has factors 3,6,7 ( for instance) can i predict what group this individual should be placed in?

Previously i have used minimum adequate models to do this. Reading through the class notes i see i should use HPDinterval(model1$Sol) however i am unsure how to interpret the results to arrive at the most significant factors for each group and simplify the model.

				lower			upper
cutpoint.group.1	0.58967173		0.85945043
cutpoint.group.2	2.31790921		2.68058107
cutpoint.group.3	3.43902338		3.84798758
cutpoint.group.4	3.91302678		4.341789
(Intercept)		0.56616994		3.61939211
a				-0.06867026		0.27986037
b				-0.14199142		0.28397164
c				-0.49411144		0.03729451
d				-0.50516484		0.08844333
e				-0.04864767		0.32516733
f				-0.12977632		0.64986726
g				-0.25288964		0.45337262
h				-0.20219456		0.26453859
i				-0.71870788		-0.16569956
attr(,"Probability")	
[1] 0.95		

Thanks and sorry for my lack of statistical knowledge!

Sam