[R-sig-Geo] Spatio-temporal data

[Barry Rowlingson]

On Sat, Jan 25, 2014 at 3:30 AM, Tirth Bhatta <tirth38 at gmail.com> wrote:
> Hi Everyone,
>
> I am currently working on a dataset that requires me to incorporate space
> and time. Space (area--census tracts) is invariant, so change in my
> variable is occurring only in terms of time. I have a count variable (count
> of disease cases) followed over time. I know I can change it to rate and
> use Gaussian models. I was trying to find out, if there are any
> spatio-temporal models that allow me to use Poisson models. I may be wrong,
> but I could only find Bayesian spatial-temporal models. Am I making a
> correct assessment? If not, what R-packages would you recommend for me? I
> saw that R-INLA may also be used, but I am not sure if I can use predictors
> in the model?

 You seem to be putting the cart before the horse. There may be lots
of spatio-temporal models (and I could probably make up some more
before lunch) but will they answer the question you want answering?

 We don't know, because you've not told us what the question is! Do
you want to know if there's spatial variability? Space-time
clustering? Correlation with a covariate? Are some areas increasing
faster than others? Etc!

 There's no point trying to find out what statistical models have been
implemented unless we know what you are trying to find out from your
data.

Barry