Hi, I'm have some data on a grid (specifically high throughput assay plates)
and am interested in evaluating measures of spatial autocorrelation to flag
plates for corrections. I have been using moran.test and geary.test from the
spdep package.
My approach is as follows:
## plate is a matrix of data
coords <- expand.grid(1:32, 1:48)
x <- as.numeric(plate)
moran.test(x, nb2listw(tri2nb(coords)))
While this works and seems to give me reasonable results, I am not
specifying a weight matrix and am imposing a triangulation on my grid.
My questions are:
1. Has anybody done something like this before, in the context of high
throughput screening?
2. I'd like to specify that for a given well on the plate, it's nearest
neighbors will the N surrounding cells. How would I go about this?
3. Is the use of tri2nb correct for this type of application?
Any pointers would be appreciated
Thanks,
--
Rajarshi Guha
NIH Chemical Genomics Center
[[alternative HTML version deleted]]
