Hi,
I am constructing a series of nonlinear mixed regression models at multiple
spatial scales on the same data. The data is a regular grid of cells. A
coarser scale is achieved, for example, by aggregating cells in blocks that
are 2x2 cells in dimension and averaging dependent and independent data over
this block. Some 2x2 blocks will be missing data for several expected
reasons and these blocks are of interest and so cannot be easily discarded
(they are also likely not at random). I would like to take this into
account when fitting the model. A simple weighting of each block by number
of complete component observations (e.g. no missing data would have a weight
of 2x2=4) seems intuitive. I've reviewed the NLME documentation and
weighting schemes seem to be the usual variety of accounting for unequal
variance. Is there a work around to specify the integer weights I described
above? I've toyed with a work around where I duplicate each block
observation by the number of observations summarized within it. Of course,
this is difficult to do correctly as the sample size will be inflated and
most statistics not easily interpretable. Any advice on how to proceed is
welcome. Thanks. -seth
--
View this message in context:
http://n4.nabble.com/weight-by-obs-in-spatial-nest-in-NLME-tp1009168p1009168.html
Sent from the R help mailing list archive at Nabble.com.
