Re: handling NaN

[John W. Eaton]

On  8-Aug-2002, Paul Kienzle <address@hidden> wrote:

| The consequences are that MISSING VALUES should be omitted by
| default.  NaNs are not missing values in many (most) computations,
| but instead are things that inform you that your computation has
| gone awry.  If your code does not explicitly deal with these
| problems itself (e.g., by using nanfun) then the invalid value
| should be propogated back to somewhere that the user can see it.

I agree with this, which is why I'm reluctant to make Octave's
functions ignore NaN by default.

| One suggestion is to define a constant NA to represent missing values.  These
| could then be ignored in all the stats functions.  I believe this would work
| well enough, except for the people coming from matlab who want to use NaN to
| represent missing values.

In R, they have NA as a special NaN value.  For some time, I thought
it might be better to have a special NA data type that would not rely
on tricks like this.  But I'm not sure that it is worth the effort.
Now, I think that having a special type means adding a lot of bloat.

So, how about adding an R-compatible NA value to Octave?  I have a
patch that does this for doubles, but that's probably all that is
needed since most things in Octave are done on doubles anyway, right?
I know that R has an NA_INTEGER value (at least internally) that is
currently represented by INT_MIN, but to me, that seems like a less
reliable thing to do, so I'd rather not if we don't need it.

In any case, with my current patch, Octave does this:

  octave:1> x = [Inf, NaN, 13, NA]    
  x =

       Inf     NaN  13.000      NA

  octave:2> isnan (x)
  ans =

    0  1  0  0

  octave:3> isinf (x)
  ans =

    1  0  0  0

  octave:4> is_NA (x)
  ans =

    0  0  0  1

  octave:5> is_NaN_or_NA (x)
  ans =

    0  1  0  1

  octave:6> finite (x)
  ans =

    0  0  1  0

  octave:7> x = rand (3); x(2,2) = NA; is_NA (x)
  ans =

    0  0  0
    0  1  0
    0  0  0

etc.

If you'd like to see the patch before I check it in, let me know (it's
about 400 lines of context diffs, so I decided it would probably be
best to omit it from this post).

Now all that's left is to fix various functions to be NA-aware.  This
will take some time, and it would help if someone (or some group)
could volunteer to help out with that.

Thanks,

jwe



-------------------------------------------------------------
Octave is freely available under the terms of the GNU GPL.

Octave's home on the web:  http://www.octave.org
How to fund new projects:  http://www.octave.org/funding.html
Subscription information:  http://www.octave.org/archive.html
-------------------------------------------------------------