[R] spatially analyzing multiple data layers

Barry Rowlingson b.rowlingson at lancaster.ac.uk
Sun May 26 01:49:19 CEST 2013

On Sat, May 25, 2013 at 7:09 PM, Amy Henry <amybethhenry at gmail.com> wrote:
> Hi, I started using R about 3 months ago so please excuse my ignorance. I
> have two datasets. The first consists of 247 fixed position mosquito traps
> that were serviced weekly for 23 weeks. The second is dengue incidence data
> indexed to home address for the same time period. I would like to see if
> there is a space time relationship between mosquito catch rate and dengue
> incidence. So far I have been unable to locate a package for multilayer
> spatial analysis or a space-time analysis that would answer the question.
> Does anyone have advise on where I should look?

 Do you have a statistical model in mind? Something like your mosquito
count data is noisy samples from an underlying spatio-temporally
smooth rate (with some space-time covariance structure), and that your
dengue incidence is dependent on that (in some way, perhaps you need
cases and controls for a logistic regression). Looks like a
spatio-temporal errors-in-variables model. Is it? Maybe you can take a
point pattern approach with the cases as an inhomogeneous space-time
poisson process with intensity given by the smoothed mosuito
measurement data... You need to figure out the stats before you go off
fishing for a package. Or are you expecting to be able to plug in your
data, crank a handle, and get an answer out?

 Have you explored the data at all? You could try some simple time
series exploratory plots (maybe there's a monthly pattern caused by
the moon?) or a few maps of the totals over all time.

 Ideally you should figure out the stats before you even collect the data...


More information about the R-help mailing list