Three steps:
First, find all the cells in each polygon, return a list of 2-column matrices with the cell number and the value:
require(plyr) # for llply, laply in a bit...
cell_value = extract(dat, polys,cellnumbers=TRUE)
head(cell_value[[1]])
cell value
[1,] 31 108
[2,] 32 108
[3,] 33 110
[4,] 92 110
[5,] 93 110
[6,] 94 111
Second, turn into a list of similar matrices but add the x and y coords:
cell_value_xy = llply(cell_value, function(x)cbind(x,xyFromCell(dat,x[,"cell"])))
head(cell_value_xy[[1]])
cell value x y
[1,] 31 108 8.581164 14.71973
[2,] 32 108 8.669893 14.71973
[3,] 33 110 8.758623 14.71973
[4,] 92 110 8.581164 14.67428
[5,] 93 110 8.669893 14.67428
[6,] 94 111 8.758623 14.67428
Third, compute the weighted mean coordinate. This neglects any edge effects and assumes all grid cells are the same size:
centr = laply(cell_value_xy, function(m){c(weighted.mean(m[,3],m[,2]), weighted.mean(m[,4],m[,2]))})
head(centr)
1 2
[1,] 8.816277 14.35309
[2,] 8.327463 14.02354
[3,] 8.993655 13.82518
[4,] 8.467312 13.71929
[5,] 9.011808 13.28719
[6,] 9.745000 13.47444
Now centr
is a 2-column matrix. In your example its very close to coordinates(polys)
so I'd make a contrived example with some extreme weights to make sure its working as expected.