http://spssx-discussion.165.s1.nabble.com/Calculating-CI-in-SPSS-when-percentage-is-close-to-0-tp5721509p5721525.html
This problem reminds me of one grave risk: What if the same var-name
exists in two files, where it does NOT denote the same variable?
This one can be nasty to discover and track down, if you don't prevent
it in the first place. (Speaking from experience.) I learned to look with
care at the MAPS provided for file merges. Most new/old variables occur
in sets.... When a variable appears by itself, ask if it belongs there.
You have text-variables? I think I would pre-process those relevant
files, before tackling the fuller set.
Maybe there is a reason that one master file is needed. However, when
this sort of question comes up, I think it is appropriate to mention the
alternatives.
I don't think I ever created one file with all the data from all the scales
and reports, though I occasionally had to work from files made that way --
I even had ones with multiple periods of data in one line.
My first step in those cases was to "normalize" the data, a concept that
I learned by reading about database management. Data that occurs together
gets its own file. Data that are parallel, like "periods", become separate lines.
Data that are essential for many analyses (age, sex, etc.) are stored together
in one well-maintained master file, which has every assigned ID, including the
ones that might be ruled out for non-participation.
Data from separate sources are "reduced" as needed, to summary totals,
composite scores, etc. Related sets - ones always used together - may be
maintained together in one file. This gives me dozens of variables to worry
about or names to scan, instead of hundreds.
For analyses: MATCH FILES brings in two or three or four files that are needed
for a particular analysis.
--
Rich Ulrich
Date: Wed, 7 Aug 2013 23:30:25 -0700
From:
[hidden email]Subject: Re: Merging multiple challenging files
To:
[hidden email]Hi,
APPLY DICTIONARY has an option /NEWVARS
that creates, well, new variables. If the variables are in the source dataset but not in the target dataset, they will be created. You could use that to build a (dummy) master file with all the dictionary info. Then you might still run into problems with string variables though.
Regards,
Albert-Jan
...snip, previous