novel use for ACEDB
JLMcCarthy at LBL.GOV
Tue Oct 12 20:30:20 EST 1999
Although I'm not currently active in the acedb community, I was a few years
ago, and I still think its a neat system. I also did a fair amount of
social science computing earlier in my career, so I'm somewhat familiar
with the kind of data you describe (mine was more for history & poli sci).
I think acedb would in fact be a good match for your data.
You would need to make your own model(s), but then could use the
query and table facilities to do most of what you describe, I think.
You might also be able to create output that you could subsequently
transform for further analysis in SAS, S, Excel, etc.
You could do the transforms with Perl, AWK, or something else if necessary.
Before too long you may be able to get the output as XML,
at which point you could use XSL-T (transformation) to do transforms.
I'd be interested to hear whether you decide to try acedb, and if so,
how it works for you.
At 01:08 PM 10/11/99 , Dave Lorand wrote:
>I'm a system administrator in the Social Science division at the
>University of Chicago, and I'm looking at ACEDB as a possible way of
>analyzing some interesting Psych data being gathered by a research
>project. I'm hoping that one of you ACEDB experts can give me a quick
>answer about whether ACEDB can handle this and how much adaptation would
>be required (I've been wading through docs for several days, and the
>learning curve is a bit steep from non-bio applications).
>This project analyzes human interaction, and their data are lists of
>actions copied down from videotapes of the interactions. For a given type
>of interaction, they have a a fairly short list of possible actions, and
>they divide a long period of interaction into short "cycles" of actions
>which turn out to fit into very neat flowcharts.
>For example, the sample data we're starting with are from a mother feeding
>a baby over a four-month period. A cycle starts when the spoon leaves the
>jar, and ends when the spoon goes back in. In between, there are a few
>places where the baby may exhibit a negative response, after which the
>mother may back off or may insist. You get the vague idea.
>The number of possible actions in these cycles is around 10, and there are
>lots of cycles.
>They want to do queries which count things like the number of cycles with
>a NEG1 (a negative response in a particular place in the flowchart) in
>which the baby was fed. Generically, they want to queries like "count
>where <action> appears before <action>", where "before" means anywhere in
>the cycle and not just immediately before.
>It seems like ACEDB could do well with these data, because cycles are
>variable-length lists of a few repeating things - very similar to gene
>sequences. Relational databases aren't so good with variable-length
>lists, and the usually Stat packages used in social science work are
>completely orthogonal to this.
>So, my question is: could ACEDB handle this easily? if not, how much
>modification and work would be required to do it?
>Thanks in advance,
>P.S. I'd appreciate email copies of any replies.
>| Dave Lorand, Programmer/Analyst | davel at src.uchicago.edu |
>| Social Science Research Computing | 773-702-3792 |
>| University of Chicago | 773-702-2101 (fax) |
John L. McCarthy Email: JLMcCarthy at lbl.gov
Building 50C -- 3318 Web: http://www.lbl.gov/~mccarthy
Berkeley National Lab Phone: (510) 486-5307
Berkeley Calif. 94720 FAX: (510) 486-4004
More information about the Acedb