```Date: Sun, 2 May 2004 23:55:21 -0400 Reply-To: Doug Fuller Sender: "SPSSX(r) Discussion" From: Doug Fuller Subject: Re: Identifying data combinations Comments: To: Frank Milthorpe Content-Type: text/plain; charset=us-ascii >>There are 2**15, or 32,768, possible combinations of >>presence and absence of 15 symptoms. That's a LONG list of >>frequencies to make sense of. Unless only a small fraction >>actually occur, and you have a number of cases that's a >>significant multiple of the number that do occur, you're >>going to have a terrible time making sense of the result. > >In determining all of these possible combinations it is >worthwhile to remember that the number of symptoms will >vary from 0 to all 15. This adds to the complexity of the >analysis. >There is no need to use the aggregate function. You can >create a single variable for each respondent (this can be >either a string or a numeric variable). Suppose there were >only three possible symptoms. The full list of possible >combinations is: [omitted] >The above can be create using either a numeric function, eg >combo = sympt1*100 + sympt2*10 + sympt3, or in a similar >manner using a string function. I can't remember what the >limit is for the maximum integer value is in SPSS, but >2**15 could be close. Once you have created this variable, >just do a frequency on the combo variable. Creating the >data in this form is relatively easy. It is what you do >with it once you have created it and how you analyse it >that is potentially much more difficult. From spssbase.pdf (v9), the Fx.y format can handle up to 40 digits or 39 digits with decimals. If space is a concern, you can certainly use smaller bases; i.e. sympt1*4 + sympt2*2 + sympt3, for the above example using base 2 representation. Of course, individual symptom retrieval is incredibly difficult visually, and takes a few lines of syntax computationally. But either solution will uniquely encode the 15 symptom configuration. -- Doug Fuller dfuller@wayne.edu Research Technician, Wayne State University Department of Psychiatry and Behavioral Neurosciences ```

Back to: Top of message | Previous page | Main SPSSX-L page