Mata: uniqrows() produces error with string matrices

[kkranker]

The documentation says that uniqrows() works with transmorphic matrix P. However, uniqrows(P,1) produces an error when P is a string matrix. For example:

sysuse surface, clear
mata:
  P = st_sdata(.,"date",.)
  uniqrows(P, 0)
  uniqrows(P, 1)
end

The source of the error is obvious from the source code (view uniqrows.mata, adopath asis ). The last line will produce an error if res is a string matrix and count is a real matrix.

 if (freq) res = res,count

Is there a way to calculate the frequencies for string matrices?

[Sergio Correia]

The problem is that a Mata matrix can't have both strings and reals, and there is no easy workaround for that.

What I do is use the ftools package from SSC (I wrote it so I'm of course biased):

mata:
x = "a" \ "a" \ "b" \ "a" \ "c"
F = _factor(x)
F.keys
F.counts
end

If you are importing the data from stata (which you are in the example), then you don't need to use "_factor()" but can use factor("date") directly

[kkranker]

Thanks - I didn't know about your ftools package, but it looks useful. I did find another word around; I mainly surprised to find that uniqrows() didn't work as documented.