| T.R | Title | User | Personal
Name | Date | Lines |
|---|
| 69.1 | A Start? | PANIC::TURNER | | Thu Jan 26 1989 16:16 | 17 |
| Peter,
I don't know anything happening in market research per se,
but there are several systems that offer help with statistics,
e.g. choosing a statistical test, interpreting results, etc.
Interestingly (well, I think it's interesting...) one technique
called multidimensional scaling is used for both market research
and knowledge acquisition. More details on request.
Another route for your query is to contact the AIIC (AI
Information Center) in the U.S.; they do searches and such.
Terry Cappuccilli and Suzanne McGuire are the contacts. Warning:
the AIIC now charges for its services; is your customer very wealthy?
Mark
|
| 69.2 | Thanks & more details | MUNSBE::BRITTAIN | | Mon Jan 30 1989 10:52 | 36 |
|
Mark,
Many thanks for the reply. The pointer to statistics
evaluation/selection packages is exactly what he needs. More details
please !
To fill in the picture a bit more: This "customer" sees Market
Research as taking place in three phases, two of which are amenable
to heuristic methods, namely the pre- and post- phases to the actual
statistic crunching.
Phase 1:(AMENABLE ?)
Selecting statistical tests, based on type and size
of market, possible samples, time of year, phase of moon,
fashions etc.
Phase 2: (not amenable ?)
(Gather and) crunch the statistics
Phase 3: (AMENABLE ?)
Evaluation of the statistics, based on validity
of sample, making allowances for
unforeseen events, data-collection gaffs etc.
I don't think I'll try the AIIC unless we do some real business
with this chap. He has just made a request for information as yet.
Still, I think his ideas are very interesting. Market research
is one of those activities which MOST if not all branches of industry
indulge in at some time or another, albeit usually via specialized
companies.
regards,
Pete.
|
| 69.3 | AI in stats update | MUNSBE::BRITTAIN | Peter, EDU Munich @UFH DTN 773-2069 | Wed Feb 01 1989 09:49 | 32 |
|
Just to keep our readers up to date...
Mark Turner has supplied me with a couple of useful references
on AI in statistics:
"Statistical Expert Systems"
by Hans. W. Gottinger (Maastricht, The Netherlands)
Expert Systems
August 1988
vol. 5, no. 3
pp. 186-196
Book "AI in Statistics"
by Gale & Pregibon (??? Publisher ???)
Taken from proceedings of the 1st conference on AI in statistics
I have also made inquiries at Bolt, Beranek & Newman (who have an
office in Munich (+49 89 225395)) with regard to their statistics
package RS1, which is reputed to have an AI component called
RS-Expert. They were very helpful and have promised to send
literature.
Watch this space!
Pete.
|
| 69.4 | BBN software | MUNSBE::BRITTAIN | Peter, EDU Munich @UFH *773-2069 | Fri Feb 10 1989 09:47 | 20 |
|
re Bolt, Beranek + Newman.
They sent me a few leaflets describing their RS series of software
products. Two of these seem to be relevant:
RS/Explore "analyses and interprets complex engineering and scientific
data. RS/Explore's four-stage approach to statistical data analysis
organises, visualises, analyses and interprets statistical results"
RS/Decision (rule based ES-tool) "...provides access to data in
RS/1 software tables as well as other comercial data management
packages. The shell allows the integration of simple rules with
essential data processing functions to maximize the payback of
developed systems."
|
| 69.5 | Statistics analysis by ES reference | MUNSBE::BRITTAIN | Peter, EDU Munich @UFH *773-2069 | Mon Feb 27 1989 09:59 | 28 |
|
Here is another reference on statistical analysis, taken from a recent
AI News Capsule. This seems to have turned into a private topic which I
am writing for posterity. This is *not* the intention. I would
appreciate any new contributions...
...anybody out there ?????
STATISTICAL ANALYSIS
++++++++++++++++++++
* "Statistical System Does the Thinking for Users"
<Digital Review FEB 6th 89 P.31>
"New Methods Research Inc has marketed a statistical analysis
program based on an ES. The Expert Statistical System is
meant for executives and researchers needing expert analysis of
raw statistics without the aid of a statistical analyst. ESS is
designed to remove the complexity customary in this type of
program. The program screens data that is entered by the user
and analyzes the data using a predetermined internal check list
of assumptions.
The program will ultimately allow user defined rules for statistical
analysis. Price: $10,000 - $30,000.
|