Because freq() depends upon tabulate(), there are a
number of workarounds in the function. Most of these are due to the way that a
vector of integers is translated into a factor object. Numeric factor objects
are formed by simply using numbers starting with 1 as levels. This
means that if zeros are the first level in a factor object, the count of zeros
will be labeled "1", the next value "2" and so on. Values that aren't present
will be skipped. So if you have a bunch of perfectly valid zeros in your data
and a category or two with no observations, you will get a result that looks
markedly different from what you might expect. Worse still, you might get a
result that looks meaningful, but doesn't represent the data in the way you
think it does.
If category labels are supplied to freq(), it uses them, but
otherwise it will try to use the entire range of values, displaying zero
observations where necessary. Zeros only seem to be a problem when they are the
first category, so the trick of adding 1 to the values is only applied in this
case. Finally, tabulate() doesn't like NAs, so they have to be
counted separately, removed, then added to the frequency counts after
tabulate() has done its job.
Finally, freq() will display category percentages if they are
requested. freq() is a good introduction to getting
R to display output in a form that can be
presented to those used to the vanilla stats report.