hmmm...thanks for the reminder, Chuck:  The person living in Alaska but working in Dallas during the "reference week" might be noting:

-- 20 minutes as his/her trip from the "regular" home in Alaska to the "regular" workplace in Alaska.

-- 20 minutes as his/her trip from the "reference week" home in Dallas to the "reference week" workplace in Dallas.

Either way, it seems that such a trip (as well as others that fit some category of strangeness) should simply be dropped from most journey-to-work regional analysis.

Which brings up a subsequent question:  Should a Planner simply drop those kinds of trips from a travel pattern analysis, or perhaps "expand" the remaining zone-to-zone trips in such a way that the original Census-estimated number of workers in the "place of work" zone is still preserved?  The good news is that the percentage of zone-to-zone trips that fit this strange category of a temporary "reference week" workplace that requires a different temporary "reference week" home place is not very large.

Ken C.

-----Original Message-----
From: Chuck Purvis [mailto:CPurvis@mtc.ca.gov]
Sent: Friday, May 21, 2004 10:42 AM
To: ctpp-news@chrispy.net
Subject: Re: [CTPP] Release of CTPP Part 3 Flat ASCII files

Hey Ken:

The ultra-long distance commuters I just throw away. I put them in the
same bucket as the weekly absentees. Or, it may help to try to sort the
commute data into: commutes from home-to-work; and, commutes from "not
home"-to-work. (Note that the census doesn't ask where you started your
commute, it asks: "at what location did this person work last week?"
Though the question on commute time departure *does* ask: "What time did
this person usually leave home to go to work last week?")

For interregional commuters, I use my common sense and select the
counties neighboring my region that have legitimate opportunities for
daily commute trips. Perhaps this is a three-hour maximum from the edge
of the region?

Counties in California are fairly large, so I use just 12 neighbor
counties surrounding the SF Bay Area. Texas has MANY smaller-sized
counties, so I imagine you would have an inner ring and outer ring(s) of
surrounding / neighboring counties. Then, if you have time, you can
build some tract-to-tract desire lines (with distance) to calculate the
tract-to-tract distances, reported average commute duration, and derived
commute speed, to determine if the interregional commutes are sensible
or silly. I don't know what I would do with a seemingly reasonable
interregional commute value that has a unreasonable travel time - -
perhaps the travel time is mis-coded, or perhaps this is another case
where the worker is NOT starting their commute from THEIR home!? I would
probably accept the numbers of commuters value, but blank out the
commute time as a missing value.

Another idea is to use the statewide travel model system to build a
statewide zone-to-zone commuter matrix, then develop commute length
(distance) frequency distributions to isolate the ultra-long,
intra-state commutes, their reported commute durations, and the derived
commute speeds. That would be preferable to a crow-flow distance desire
line analysis.

(By the way, I haven't tested any of these methods. They're just some
ideas to share.)

Chuck Purvis, MTC

>>> Ken Cervenka <kcervenka@nctcog.org> 05/18/04 03:56PM >>>
This question has probably been asked before, but I don't recall the
answer...

What sort of adjustments do people make with CTPP data, to deal with
the
extreme "long-distance" commuters that are actually just "temporarily
re-assigned" workers?  This is not an error in the CTPP collection or
summary process, but rather a result of the way the long form questions
were
answered.

For example, here is a record that got "expanded" to four people:
-- The Place of Residence is a County in Alaska
-- The Place of Work is a County in Texas (Dallas)
-- The commute time is 20 minutes

What this is saying is that the person isn't really commuting each day
from
their home in Alaska, but is instead commuting to a place in Dallas
County
from some temporary residence (a hotel, perhaps) within the
Dallas-Fort
Worth area.

Such an extreme example is easy to recognize, but what about other
examples
which fall into the "not sure" category when it comes to whether a
person's
normal place of residence is really the point where they are truly
starting
their commute trip?  Perhaps a very high "average speed" based on
"travel
time" and "straight-line distance" could be used to flag such records?

Ken Cervenka
North Central Texas COG