[R] eof and svd calculus with NA's

[(Ted Harding)]

On 12-Sep-03 antonio rodriguez wrote:
> Im currently dealing with large datasets of some climatic parameters
> and I'm performing some EOF analysis on them. The problem is that for
> one of the datasets, the continents are labelled as NA's (since the
> data was gathered over the oceans). I don't know to which extent the
> dropping of those NA's from the matrix will affect the future
> calculus, and how to maintain the geographic entity of the EOF's

Surely you can sidestep any potential problems of dropping data by
a global editing of this dataset so that all "NA" continents become,
say, "Ocean"?

Not much more informative than "NA" perhaps, but still more meaningful
and will avoid the problem you raise.

Best wishes,
Ted.