[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]
RE: Missing values [was: RE: st: simple question]
Is it worth pointing out that SAS, which may be the most widely used stat
package on the planet, uses missing value codes almost exactly the same way that
Stata uses them. S-Plus uses an internal code (similar to old-style Stata's
".") and LIMDEP (at least at Version 6.0) used -999 to indicate missing. While
I too find it sometimes convenient to define arbitrary values as missing in
SPSS, this by no means is a universal approach and has even led to subtle errors
when the missing codes were not properly defined (or handled) as missing. (Of
course, I never made such errors, but I've heard plenty of stories....)
Nevertheless, I personally prefer the SAS/Stata approach over the SPSS approach.
I also agree that Stat/Transfer should endeavor to preserve unique missing value
codes whenever possible, if it does not do so.
> -----Original Message-----
> From: firstname.lastname@example.org
> [mailto:email@example.com]On Behalf Of Richard
> Sent: Wednesday, June 02, 2004 2:41 PM
> To: firstname.lastname@example.org
> Subject: Re: Missing values [was: RE: st: simple question]
> At 10:18 PM 6/2/2004 +0100, Nick Cox wrote:
> >Perhaps you could expand on this.
> Sure. What I dislike is that, unlike other programs, Stata forces you to
> recode the MD values from a data set. All those 7s, 9s, 99s, etc. will
> have to be recoded to .a, .b, etc. The values you use for MD will not be
> the same as the values in the published codebooks or that may be reported
> by other users of the data who are using other programs.
> In SPSS, on the other hand, I would just say something like
> Missing Values X (77, 88, 99).
> This isn't a mega problem, but when Stata expanded its possible MD codes
> I'd have rather seen it take an SPSS-type approach than the route it did
> go. (On the other hand, I suppose you could argue that the rest of the
> world should do things more like Stata.)
> The bigger problem (which is perhaps more of an issue with Stat/Transfer
> than it is with Stata) is that distinctions between types of MD codes get
> lost when converting from SPSS to Stata. I don't know whether this problem
> is unique to Stata or occurs with other types of conversions as well, but I
> would like to see software better address it, e.g. when there are multiple
> MD codes in SPSS assign them to .a, .b, etc. in Stata.
> * For searches and help try:
> * http://www.stata.com/support/faqs/res/findit.html
> * http://www.stata.com/support/statalist/faq
> * http://www.ats.ucla.edu/stat/stata/
* For searches and help try: