[R] (no subject)

[Nick Wray]

Hello - I am using the MICE package to impute missing values - the input
data are up to six vectors of parallel hourly temperature measurements for
Scottish weather stations across a calendar year. None of the vectors have
more than 5% NAs as I have filtered ones with more out.  Most of the sets
work fine with MICE but with a few I get an error message:

This data set, which generated the error message has five columns

iter imp variable
  1   1  986Error in terms.formula(tmp, simplify = TRUE) :
  invalid term in model formula


986 is the station number which is the column name here for the first
column.  The third to fifth columns don't have any NAs and the first and
second have fewer than 1% - but they do have the NAs concentrated as
strings of 20 or so near the beginning of the data set.

I am puzzled as to why certain datasets, which don't seem very different
from the others provoke the error message - the only thing which I am
wondering is whether MICE has a problem with a too large concentration of
the NAs in particular regions but I can't find any reference to this in the
literature.  Has anyone else come across this as a problem, and if so, what
did they do about it?  Thanks Nick Wray

	[[alternative HTML version deleted]]