On Apr 2, 10:55*am, Ron Wanttaja wrote:
On Wed, 02 Apr 2008 08:24:07 -0500, Gig 601Xl Builder

wrote:
Ron Wanttaja wrote:
Rollicking good time in the Wanttaja household lately, with another set of
homebuilt registrations statistics/homebuilt accident statistics analyses under
way.

interesting data there Ron. RVs make up 14.2% of the fleet. That is
impressive.

How are you pulling the data for those folks that don't keep the
designer's name on the registration in any way?

No way of telling if they call their RV-6 a "Jones Special," of course. *Most
keep *some* sort of relationship with the designer/kit maker's designation for
the aircraft.

When developing filters, I start with something fairly broad, then examine the
hits to discover what other designations might be used for the aircraft, then
tune from there.

For example, when trying to spot the Glastars, I initially ran a search on just
that name. *Looking at the listings, I then see that a numeric designation of
the design can be "GS-1" or "SH-4". *So I add variations of them to the
filter... I look for "GS-1", "GS 1", and "GS1", and the same variations on SH-4.

Unfortunately, when I search for "*GS 1*", it ALSO hits the Schweizer SGS 1-23
glider. *So I have to add exclusions for similar but incorrect designations.
Sometimes I have to add a filter just to keep one sneaky incorrect entry from
popping up.

Here are my ultimate criteria for finding the Glastars:

(Like "*SH-4*" Or Like "*SH4*" Or Like "*SH 4*") And Not Like "*CONTROL*" And
Not Like "*Sportsman*"

or

(Like "*SH-4*" Or Like "*SH4*" Or Like "*SH 4*") And Not Like "*CONTROL*" And
Not Like "*Sportsman*"

I exclude "Sportsman" from my Glastar search because I have a separate filter
for the Sportsman (which, in itself, includes filters to eliminate the Wag-Aero
Sportsman and other planes using that name). *I eventually had to include a
year-of-manufacture filter just to get rid of some of the non-Glastar variants.

I've seen listings where they *spell* out the number (e.g. "BD FIVE"), but
they're rare enough that I don't routinely add that kind of filter. *The
filtration is never perfect, of course. *I don't catch the misspellings
("Glassair") or name abbreviations ("STRD HMLTN SUPER IIS") and other such
variations. *In my experience, though, the variations are rare enough that they
don't really impact the numbers much.

The list I posted accounts for about *half* the total number of Experimental
Amateur-Built registrations. *The remaining types are a wide variety...
Breezies, CA-65s, Stephens, Miniplanes, Jungsters, Cassutts, Sonerais, etc.. *A
few dozen of one type, a hundred or so of another, it adds up. *If anyone feels
I left a major type off my list, let me know.

Ron Wanttaja

Maybe this will help you in classifying entries. In real life my job
is database programming and data management. In situations like this
I'll put the data into a temporary table with an additional column
called "Processed" I'll run my statements wherein I am trying to find
a particular pattern and when I get a good match, e.g. Schweizer SGS
1-23, I would mark it as processed. When I later try to match the GS
1 entries I can exclude the ones that were already processed so that
my pattern that is a little broader than I'd like has most of the
potential bad matches already moved from its source list.