Advertisement
Anyone out there using BUGS?
I'm just starting out...and could do with links to tutorial / introductions with a bit more explanation that the included examples.
Can anyone help, please?
I'm just starting out...and could do with links to tutorial / introductions with a bit more explanation that the included examples.
Can anyone help, please?
posted by:
|
Unsubscribed |
Advertisement
Advertisement
-
Unsu...
Re: BUGS beginner
Wed, September 28, 2005 - 9:57 AMnope, i don't use BUGS, tho people have suggested it to me. i find it easier to write my own software than to learn other peoples', but that's just me.
most numerical methods are pretty easy to implement yourself in, say, R. -
-
Unsu...
Re: BUGS beginner
Wed, September 28, 2005 - 3:50 PMHi Bobby, new friend of mine!
Mostly, I code in R & C. But for this, I don't have time to code and don't want to have ot reinvent the wheel. Unfortunately, I just got the larger data set - I'd been working with a trial data set before...and its huge so BUGs probably wouldn't be capable of handling it. Still it'd be nice to try this stuff out before spending a lot of time coding.
So, here's my problem. I have marine species data for 8881 species (Y) and 32 spatial covariates (X). I want to essentially cluster the species into groups (C) and predict those groups using the spatial covariates. I was thinking of a latent variable model with a dag that looks something like this:
X -> C -> Y
Have you seen any R/C/C++ code to do anything like that?
Cheers, Fi (aka ghostcat). -
-
Unsu...
Re: BUGS beginner
Wed, September 28, 2005 - 4:14 PMwould your DAG be like a hidden markov model? also, what does your data look like? is there a vector of spatial covariates for each species? if so, there are a number of clustering algorithms that could work, tho i haven't seen/used any of these in R, C, etc. there is the program Cluster from michael eisen's group, used for gene expression data, but can be used to cluster pretty much anything where the data is in the form of vectors.
you could try the book 'finding groups in data' by kaufman and rousseeuw if you decide to roll your own algorithm. -
-
Unsu...
Re: BUGS beginner
Wed, September 28, 2005 - 7:38 PMNot an HMM - more like a Bayesian network where the latent class is conditioned on the spatial covariates (a vector in R^32) and the species (a vector in R^8881) are conditioned on the class. -
-
Unsu...
Re: BUGS beginner
Thu, September 29, 2005 - 10:28 AMi'm stumped, other than trying existing clustering algorithms. the cluster package in R might be useful..?
-
-
-
-
