At 02:09 AM 6/22/2005, Omar Farook wrote:

>I am searching to know the use of the Draft Output. Any comments, >notes and references I appreciate that.

At 10:28 AM 6/22/2005, John Norton responded:

>The Draft Output is usually used by what I used to refer to in the >Basics Training classes as power users and those not immediately >concerned with producing publishable quality output items like pivot >tables and high resolution graphics. Essentially, the contents of the >Draft Viewer are written using text characters, so tables use a series >of dashes and "pipe" characters for lines, and plus signs for where >those lines intersect.

Exactly. In particular, - Any complicated transformation program, I always run to draft output. I'm still not clear whether it's merely difficult, or literally impossible, to see a long stretch of code in standard output. I always see the beginning, only - When I need output that can be treated as text. As one example, I always run test runs to draft output for copying and pasting into postings to this list. - For simple frequencies and crosstabs, sometimes means, to get a quick check on something in the data. It's not as pretty as pivot-table output, but it's quicker to get to, once it's run. (It's also quicker to run, but that's because I start with a draft-output window open by default.)

John also wrote,

>[Draft output] it reduces the demand on system resources for memory >intensive operations, and can actually be a viable option when >increased speed is a concern.

That's a new one on me, but I don't use regular output often. It make sense that pivot-table output takes more more processing time, but 'more' of what's a small proportion of the total: preparing the presentable output, after the statistical work is done.

Commonly, the time-consuming parts of a run are, A. Running through the data file to gather the necessary summary statistics. (This is strongly dependent on the size of the file being processed.) B. Processing the summary statistics into the final result. (This can be quite complex for some procedures, like factor analysis, that do a lot of math on the 'summary statistics', which are basically a correlation matrix.)

If draft output is noticeably faster, I'd suppose it would be when you're running a lot of mathematically simple procedures against a small data file. John, is this when you see noticeable improvements? Other times as well?