Specifies a control file to load in, specifying
filenames and binary classifications for the texts. The file should
contain be three sep-separated (or whitespace-delimited, if sep is
NULL as in the default case) columns, one headed ``filename''
providing a list of filenames, one headed ``truth'' providing the
classifications for a subset and missing values (NA or ``.'' for the
others), and a third headed ``trainingset'' and having a 1 for each
element of the training set and a 0 for elements of the test set.
When trainingset=1, truth should not be missing or
it will be deleted. The function will compute the distribution of
documents across categories for all documents with
trainingset=0 (if truth is not missing for some
these observations, it will not be used during estimation but will
be used for printing and graphics on output to compare to the
estimates). Defaults to ``control.txt''.
Alternatively, one can provide a data frame in the same three-column
format. This will be written to readmetmpctrl.txt in the working
directory during program operation.