Developing a "Transit Score" for all locations

86 views
Skip to first unread message

Brandon Martin-Anderson

unread,
Nov 3, 2009, 3:15:42 PM11/3/09
to transit-d...@googlegroups.com
Hey transit dev-folks.

I'm working with Front Seat Software on a project called "Walk Score", which rates the walkability of every address in the US on a scale of 0 to 100. The idea is simple and brilliant - if the walkability of every location is consistently rated then real estate agents can list the walk scores along with properties, providing incentive for the real estate industry to support walkability, in addition to a way to educate and increase awareness of walkability in general.

One of the most requested new features of Walk Score is the inclusion of public transit into the score. The complexity of public transit and the non-uniformity of transit data actually makes this a pretty tricky issue. How valuable is a stop that's 1/4 mile away? Compared to 1/2 mile? What about bus vs. train? Is 40 trips per day twice as valuable as 20 trips per day? Is a transit station with 40 routes 40 times more valuable than a transit station with a single route?

At Front Seat we've been documenting our thoughts on a wiki page here:

http://wiki.walkscore.org/transit-score

I'd be really interested to hear what anybody has to say about how they would build a transit score, or what you think of the ways we're considering. Specifically we've been debating the function relating amenity value to amenity distance. Sigmoidal? Inverse proportional? Linear? What's the conventional thinking on this?

If you're interested in contributing, please create an account at walkscore.org (or use a Google account) and leave a comment on the transit-score page, or email me.

Cheers,
B

fruminator

unread,
Nov 3, 2009, 3:46:57 PM11/3/09
to Transit Developers
Brandon,

What you're basically talking about is well-known as a general concept
called 'accessibility' which is sort of a reaction to an obsession
with 'mobility.'

These 2 theses should give a decent review and set of references on
the topic of accessibility:

http://dspace.mit.edu/handle/1721.1/32414 "Accessibility-based transit
planning"
http://dspace.mit.edu/handle/1721.1/40130 "The accessibility and
development impacts of new transit infrastructure : the Circle Line in
Chicago"

Mike

Jerry J

unread,
Nov 4, 2009, 12:40:48 PM11/4/09
to Transit Developers
fruminator gave an excellent response with the concept of
accessibility and those sources. Many researchers in this field have
translated the metric "how far you can go in 45 minutes" to the number
of job opportunities or medical facilities (destinations relevant to a
transit rider) that can be reached in this time.

check out this paper by Robert Cervero: Tracking accessibility:
employment and housing opportunities in the San Francisco Bay Area
http://envplan.com/abstract.cgi?id=a311259
At Berkeley he is the professor that has done the most research in
this field (transportation and land use)

Also, I haven't read this one but it seems relevant: Measuring
accessibility: an exploration of issues and alternatives
http://www.envplan.com/abstract.cgi?id=a291175

I'm not at school right now, so I don't have access to the PDFs, but
if you are interested in taking a look, I can download them and send
them to you in a couple hours. Also, maybe a city planner person can
chime in about this also.

Also, I guess I should mention that I am working on something
related. Transitshed creates travel time maps based on schedule
information. From GPS traces of buses, I am also creating travel time
maps, and trying to find areas of major discrepancy between the
schedule and actual performance of the transit system and maybe the
reasons for these problems. I only have the data now for the sf bay
area (thanks to NextBus) but maybe this can help out with the transit
score in the future?

Jerry

Tommy B Goode

unread,
Nov 4, 2009, 12:44:12 PM11/4/09
to transit-d...@googlegroups.com
For Philadelphia your calculations are very simple at the moment. (:
--
Tom Boutell

http://www.tommybgoode.com/
http://boutell.livejournal.com/

Philip Ashlock

unread,
Nov 4, 2009, 2:02:20 PM11/4/09
to Transit Developers
I'm sure some of you have seen this before, but http://www.triptropnyc.com/
provides a very simple visualization of accessibility based on subway
metrics. Simultaneous correlation with buses would make this even more
accurate.

-phil

fruminator

unread,
Nov 4, 2009, 2:36:15 PM11/4/09
to Transit Developers
these are 'isochrones'.

isochones = accessibility only IF you assume the attractiveness of
destinations falls of linearly with travel time. This is likely not
the case, thus Brandon's question about the shape of the function.

as for Jerry J, the comparison of actual to scheduled running times is
a big deal so I'm glad to hear you are doing that. Rarely is that
sort of thing aired publicly. There are a lot of reasons that there
will be discrepancies, some good and some bad. But to get you
started, let me ask this question -- assume that you had the scheduled
running time and you measure a distribution of running times, then how
would you set the scheduled running time as a function of that
distribution? the min() ? the max() ? the mean()? the median() ? the
95th percentile? etc. this gets into a fairly complicated set of
tradeoffs between cost and reliability and passenger wait times and
passenger in-vehicle travel times and layover time at terminals and en-
route holding policies etc.

Mike

On Nov 4, 2:02 pm, Philip Ashlock <p...@openplans.org> wrote:
> I'm sure some of you have seen this before, buthttp://www.triptropnyc.com/

Kieran Huggins

unread,
Nov 4, 2009, 4:20:36 PM11/4/09
to transit-d...@googlegroups.com
Since you're looking for a _general_ accessibility score, I think it would be best to calculate the stoptime frequency graph against a general ridership volume graph for the city in question. For example: one bus 20 minutes apart during rush hour is *less* valuable than the same bus 30 minutes apart in the middle of the night. Does that make sense?

Might also be sage to take the direction of the service into account as well, thereby calculating a total "value area" as a combination of vectors, whose length is proportional to the frequency. Thus, an intersection with buses in multiple directions would be considerably more valuable, while hubs and stations would have the highest value. You could also factor the average speed of each vehicle into that equation (subways > express buses > buses).

Without having more information about the specific visitor (like where they want to go and when), I'm not sure what else can really be done.

Cheers,
Kieran

T Sobota

unread,
Nov 5, 2009, 2:36:01 PM11/5/09
to Transit Developers
Semantically as it relates to buses and bus stops, I would think one
should clarify that accessiblity is related to (unique) scheduled
trips that serve a bus stop (or stops).

For an address in the "center" of a neighborhood, where the individual
could just as well walk two blocks west to a bus stop bus to catch the
bus - or two blocks north to a different bus stop location to catch
the same bus (the route making an eastbound right turn from one
arterial to the other), the fact of being "served" by two bus stops
only equates to being served by one bus (route) or however many
scheduled trips. Conversely, the north-south arterial may have only
the branch of one scheduled route direction - while the east-west
arterial could be carrying a trunk of trips from both that route
direction as well as additional trips on another route direction/bus
line.

I'd also note that trips per (typical) week would seem to best
encompass the various daily schedule deviations that might befall a
bus stop location (i.e. weekday, saturday, sunday service, extra trips
late night on a friday only, etc.)
> > > >http://www.tommybgoode.com/http://boutell.livejournal.com/- Hide quoted text -
>
> - Show quoted text -

Aaron Antrim

unread,
Nov 29, 2009, 2:24:41 AM11/29/09
to Transit Developers, Jarrett Walker
I am most interested in something along the lines of: "4.3 Distance-
penalized trip-weighted (calculating the Walk Score at all stops along
the way) service-level summing."

With that proposal (4.3), it does look as if nearby stops on the same
line may be counted, even though those access points wouldn't offer
additional value (as Tim Sobota pointed out). Maybe this is jumping
in the weeds sooner than necessary -- I assume this would get worked
out later. Jarret Walker (humantransit.org) points out that "transit
planners are often trying to take stops out and/or implement policies
that create consistent and generally wider spacing" (http://
tinyurl.com/yesz6h8). Weighting the transit score by the number of
stops in the area won't be very useful. Additionally, the
distance_penalty shouldn't be so great that it significantly reduces
transit scores in the region where the system has widened stop
spacing.

I'm not a fan of the idea of weighting the trip values differently
according to mode — rail vs. bus, for example. The "distance-
penalized trip-weighted…" approach will show if there is a difference
in service speeds that makes more of the city accessible, anyway. I
acknowledge that there are differences in ride quality between
different types of vehicles. But the distinction between bus and rail
in terms of service quality, station amenities, (and who each is for)
etc. should, and is, blurring. If people really want to be near light
rail, the alignments and stations are usually pretty easy to find even
without a transitscore.

-Aaron
http://www.trilliumtransit.com/blog
> > > > >http://www.tommybgoode.com/http://boutell.livejournal.com/-Hide quoted text -

Brandon Martin-Anderson

unread,
Nov 30, 2009, 1:40:30 PM11/30/09
to transit-d...@googlegroups.com
It's interesting you should say that you don't think buses have a
different impedance than trains. We've estimated that the impedance of
rail travel is nearly half that of bus travel by virtue of the mode
alone. That's just our intuition, however, and I'd be really curious
to see any literature on the subject.

-B
> --
>
> You received this message because you are subscribed to the Google Groups "Transit Developers" group.
> To post to this group, send email to transit-d...@googlegroups.com.
> To unsubscribe from this group, send email to transit-develop...@googlegroups.com.
> For more options, visit this group at http://groups.google.com/group/transit-developers?hl=en.
>
>
>

Matt Laird

unread,
Nov 30, 2009, 2:27:57 PM11/30/09
to transit-d...@googlegroups.com
Unfortunately I don't have any links off the top of my head but I've
heard similar studies quoted before, that there is a definite
preference for rail over bus in attractiveness to riders.

Based on intuition I'd tend to agree with this as well. There's just
something more attractive about a rail vehicle with a feeling of more
space, typically a smoother ride because you're not starting and
stopping in traffic and are on smooth rails instead of worn roads, and
the feeling of a faster speed due to right-of-way separation.

Here in Vancouver there's definitely anecdotal evidence that
development and housing prices seem to increase closer to rapid
transit stations when it comes to multi-family housing developments.
So for any transit score, a differentiation between bus and rail would
certainly seem appropriate.

Michael Frumin

unread,
Nov 30, 2009, 2:40:16 PM11/30/09
to transit-d...@googlegroups.com
There is infinite empirical research on the subject, and of course none of it agrees.

One place to start is with the brits, since they love cost-benefit analysis, which depend heavily on relative values-of-time.� eg:
http://eprints.whiterose.ac.uk/3393/
http://eprints.whiterose.ac.uk/2062/
http://eprints.whiterose.ac.uk/2086/

My general feeling having read and skimmed some of this literature is that after you control for everything else -- headway, travel time, reliability, crowding, etc -- the inherent disutility from riding in a bus isn't that different from a train, at least for short-ish urban trips.

Mike

Matt Laird wrote:
I am most interested in something along the lines of: "4.3 Distance-
penalized trip-weighted (calculating the Walk Score at all stops along
the way) service-level summing."

With that proposal (4.3), it does look as if nearby stops on the same
line may be counted, even though those access points wouldn't offer
additional value (as Tim Sobota pointed out). �Maybe this is jumping
in the weeds sooner than necessary -- I assume this would get worked
out later. �Jarret Walker (humantransit.org) points out that "transit
planners are often trying to take stops out and/or implement policies
that create consistent and generally wider spacing" (http://
tinyurl.com/yesz6h8). �Weighting the transit score by the number of
stops in the area won't be very useful. �Additionally, the
distance_penalty shouldn't be so great that it significantly reduces
transit scores in the region where the system has widened stop
spacing.

I'm not a fan of the idea of weighting the trip values differently
according to mode � rail vs. bus, for example. �The "distance-
penalized trip-weighted�" approach will show if there is a difference
in service speeds that makes more of the city accessible, anyway. �I
acknowledge that there are differences in ride quality between
different types of vehicles. �But the distinction between bus and rail
in terms of service quality, station amenities, (and who each is for)
etc. should, and is, blurring. �If people really want to be near light
rail, the alignments and stations are usually pretty easy to find even
without a transitscore.

-Aaron
http://www.trilliumtransit.com/blog

On Nov 5, 11:36�am, T Sobota <tsob...@cityofmadison.com> wrote:
      

      
Semantically as it relates to buses and bus stops, I would think one
should clarify that accessiblity is related to (unique) scheduled
trips that serve a bus stop (or stops).

For an address in the "center" of a neighborhood, where the individual
could just as well walk two blocks west to a bus stop bus to catch the
bus - or two blocks north to a different bus stop location to catch
the same bus (the route making an eastbound right turn from one
arterial to the other), the fact of being "served" by two bus stops
only equates to being served by one bus (route) or however many
scheduled trips. �Conversely, the north-south arterial may have only
the branch of one scheduled route direction - while the east-west
arterial could be carrying a trunk of trips from both that route
direction as well as additional trips on another route direction/bus
line.

I'd also note that trips per (typical) week would seem to best
encompass the various daily schedule deviations that might befall a
bus stop location (i.e. weekday, saturday, sunday service, extra trips
late night on a friday only, etc.)

On Nov 4, 3:20�pm, Kieran Huggins <kie...@kieran.ca> wrote:

        
isochones = accessibility only IF you assume the attractiveness of
destinations falls of linearly with travel time. �This is likely not
the case, thus Brandon's question about the shape of the function.
            
as for Jerry J, the comparison of actual to scheduled running times is
a big deal so I'm glad to hear you are doing that. �Rarely is that
sort of thing aired publicly. �There are a lot of reasons that there
will be discrepancies, some good and some bad. �But to get you
started, let me ask this question -- assume that you had the scheduled
running time and you measure a distribution of running times, then how
would you set the scheduled running time as a function of that
distribution? �the min() ? the max() ? the mean()? the median() ? the
95th percentile? etc. �this gets into a fairly complicated set of
tradeoffs between cost and reliability and passenger wait times and
passenger in-vehicle travel times and layover time at terminals and en-
route holding policies etc.
            
Mike
            
On Nov 4, 2:02 pm, Philip Ashlock <p...@openplans.org> wrote:
            
I'm sure some of you have seen this before, buthttp://
              
www.triptropnyc.com/
            
provides a very simple visualization of accessibility based on subway
metrics. Simultaneous correlation with buses would make this even more
accurate.
              
-phil
              
On Nov 4, 12:44 pm, Tommy B Goode <tommybgo...@gmail.com> wrote:
              
For Philadelphia your calculations are very simple at the moment. (:
                
On Wed, Nov 4, 2009 at 12:40 PM, Jerry J <jjari...@gmail.com> wrote:
                
fruminator gave an excellent response with the concept of
accessibility and those sources. �Many researchers in this field have
translated the metric "how far you can go in 45 minutes" to the
                  
number
            
of job opportunities or medical facilities (destinations relevant to
                  
a
            
transit rider) that can be reached in this time.
                  
check out this paper by Robert Cervero: �Tracking accessibility:
employment and housing opportunities in the San Francisco Bay Area
http://envplan.com/abstract.cgi?id=a311259
At Berkeley he is the professor that has done the most research in
this field (transportation and land use)
                  
Also, I haven't read this one but it seems relevant: Measuring
accessibility: an exploration of issues and alternatives
http://www.envplan.com/abstract.cgi?id=a291175
                  
I'm not at school right now, so I don't have access to the PDFs, but
if you are interested in taking a look, I can download them and send
them to you in a couple hours. �Also, maybe a city planner person can
chime in about this also.
                  
Also, I guess I should mention that I am working on something
related. �Transitshed creates travel time maps based on schedule
information. �From GPS traces of buses, I am also creating travel
                  
time
            
maps, and trying to find areas of major discrepancy between the
schedule and actual performance of the transit system and maybe the
reasons for these problems. �I only have the data now for the sf bay
area (thanks to NextBus) but maybe this can help out with the transit
score in the future?
                  

Brandon Martin-Anderson

unread,
Nov 30, 2009, 2:47:53 PM11/30/09
to transit-d...@googlegroups.com
I also suspect that if your value analysis controls for all the things
that distinguish a train from a bus then the value of a train would be
identical to a bus. :)

On Mon, Nov 30, 2009 at 11:40 AM, Michael Frumin <mic...@frumin.net> wrote:
> There is infinite empirical research on the subject, and of course none of
> it agrees.
>
> One place to start is with the brits, since they love cost-benefit analysis,
> which depend heavily on relative values-of-time.  eg:
> additional value (as Tim Sobota pointed out).  Maybe this is jumping
> in the weeds sooner than necessary -- I assume this would get worked
> out later.  Jarret Walker (humantransit.org) points out that "transit
> planners are often trying to take stops out and/or implement policies
> that create consistent and generally wider spacing" (http://
> tinyurl.com/yesz6h8).  Weighting the transit score by the number of
> stops in the area won't be very useful.  Additionally, the
> distance_penalty shouldn't be so great that it significantly reduces
> transit scores in the region where the system has widened stop
> spacing.
>
> I'm not a fan of the idea of weighting the trip values differently
> according to mode — rail vs. bus, for example.  The "distance-
> penalized trip-weighted…" approach will show if there is a difference
> in service speeds that makes more of the city accessible, anyway.  I
> acknowledge that there are differences in ride quality between
> different types of vehicles.  But the distinction between bus and rail
> in terms of service quality, station amenities, (and who each is for)
> etc. should, and is, blurring.  If people really want to be near light
> rail, the alignments and stations are usually pretty easy to find even
> without a transitscore.
>
> -Aaron
> http://www.trilliumtransit.com/blog
>
> On Nov 5, 11:36 am, T Sobota <tsob...@cityofmadison.com> wrote:
>
>
> Semantically as it relates to buses and bus stops, I would think one
> should clarify that accessiblity is related to (unique) scheduled
> trips that serve a bus stop (or stops).
>
> For an address in the "center" of a neighborhood, where the individual
> could just as well walk two blocks west to a bus stop bus to catch the
> bus - or two blocks north to a different bus stop location to catch
> the same bus (the route making an eastbound right turn from one
> arterial to the other), the fact of being "served" by two bus stops
> only equates to being served by one bus (route) or however many
> scheduled trips.  Conversely, the north-south arterial may have only
> the branch of one scheduled route direction - while the east-west
> arterial could be carrying a trunk of trips from both that route
> direction as well as additional trips on another route direction/bus
> line.
>
> I'd also note that trips per (typical) week would seem to best
> encompass the various daily schedule deviations that might befall a
> bus stop location (i.e. weekday, saturday, sunday service, extra trips
> late night on a friday only, etc.)
>
> On Nov 4, 3:20 pm, Kieran Huggins <kie...@kieran.ca> wrote:
>
>
>
> Since you're looking for a _general_ accessibility score, I think it would
> be best to calculate the stoptime frequency graph against a general
> ridership volume graph for the city in question. For example: one bus 20
> minutes apart during rush hour is *less* valuable than the same bus 30
> minutes apart in the middle of the night. Does that make sense?
>
>
> Might also be sage to take the direction of the service into account as
> well, thereby calculating a total "value area" as a combination of vectors,
> whose length is proportional to the frequency. Thus, an intersection with
> buses in multiple directions would be considerably more valuable, while hubs
> and stations would have the highest value. You could also factor the average
> speed of each vehicle into that equation (subways > express buses > buses).
>
>
> Without having more information about the specific visitor (like where they
> want to go and when), I'm not sure what else can really be done.
>
>
> Cheers,
> Kieran
>
>
> On Wed, Nov 4, 2009 at 2:36 PM, fruminator <mich...@frumin.net> wrote:
>
>
> these are 'isochrones'.
>
>
> isochones = accessibility only IF you assume the attractiveness of
> destinations falls of linearly with travel time.  This is likely not
> the case, thus Brandon's question about the shape of the function.
>
>
> as for Jerry J, the comparison of actual to scheduled running times is
> a big deal so I'm glad to hear you are doing that.  Rarely is that
> sort of thing aired publicly.  There are a lot of reasons that there
> will be discrepancies, some good and some bad.  But to get you
> started, let me ask this question -- assume that you had the scheduled
> running time and you measure a distribution of running times, then how
> would you set the scheduled running time as a function of that
> distribution?  the min() ? the max() ? the mean()? the median() ? the
> 95th percentile? etc.  this gets into a fairly complicated set of
> tradeoffs between cost and reliability and passenger wait times and
> passenger in-vehicle travel times and layover time at terminals and en-
> route holding policies etc.
>
>
> Mike
>
>
> On Nov 4, 2:02 pm, Philip Ashlock <p...@openplans.org> wrote:
>
>
> I'm sure some of you have seen this before, buthttp://
>
>
> www.triptropnyc.com/
>
>
> provides a very simple visualization of accessibility based on subway
> metrics. Simultaneous correlation with buses would make this even more
> accurate.
>
>
> -phil
>
>
> On Nov 4, 12:44 pm, Tommy B Goode <tommybgo...@gmail.com> wrote:
>
>
> For Philadelphia your calculations are very simple at the moment. (:
>
>
> On Wed, Nov 4, 2009 at 12:40 PM, Jerry J <jjari...@gmail.com> wrote:
>
>
> fruminator gave an excellent response with the concept of
> accessibility and those sources.  Many researchers in this field have
> translated the metric "how far you can go in 45 minutes" to the
>
>
> number
>
>
> of job opportunities or medical facilities (destinations relevant to
>
>
> a
>
>
> transit rider) that can be reached in this time.
>
>
> check out this paper by Robert Cervero:  Tracking accessibility:
> employment and housing opportunities in the San Francisco Bay Area
> http://envplan.com/abstract.cgi?id=a311259
> At Berkeley he is the professor that has done the most research in
> this field (transportation and land use)
>
>
> Also, I haven't read this one but it seems relevant: Measuring
> accessibility: an exploration of issues and alternatives
> http://www.envplan.com/abstract.cgi?id=a291175
>
>
> I'm not at school right now, so I don't have access to the PDFs, but
> if you are interested in taking a look, I can download them and send
> them to you in a couple hours.  Also, maybe a city planner person can
> chime in about this also.
>
>
> Also, I guess I should mention that I am working on something
> related.  Transitshed creates travel time maps based on schedule
> information.  From GPS traces of buses, I am also creating travel
>
>
> time
>
>
> maps, and trying to find areas of major discrepancy between the
> schedule and actual performance of the transit system and maybe the
> reasons for these problems.  I only have the data now for the sf bay

Michael Frumin

unread,
Nov 30, 2009, 2:51:59 PM11/30/09
to transit-d...@googlegroups.com
I can't tell if that's a sarcastic paraphrasing.� I didn't mean to say that these studies control for _all_ things.� For example I'm sure they were not able to find any steel-wheel-on-rail buses or rubber-tire-on-road trains.

What this is useful for is to help focus attention on what matters -- protecting transit from private vehicle congestion, improving reliability and frequency, etc.

Brandon Martin-Anderson wrote:
I also suspect that if your value analysis controls for all the things
that distinguish a train from a bus then the value of a train would be
identical to a bus. :)

On Mon, Nov 30, 2009 at 11:40 AM, Michael Frumin <mic...@frumin.net> wrote:
  
There is infinite empirical research on the subject, and of course none of
it agrees.

One place to start is with the brits, since they love cost-benefit analysis,
which depend heavily on relative values-of-time.� eg:
 wrote:


I am most interested in something along the lines of: "4.3 Distance-
penalized trip-weighted (calculating the Walk Score at all stops along
the way) service-level summing."

With that proposal (4.3), it does look as if nearby stops on the same
line may be counted, even though those access points wouldn't offer
additional value (as Tim Sobota pointed out). �Maybe this is jumping
in the weeds sooner than necessary -- I assume this would get worked
out later. �Jarret Walker (humantransit.org) points out that "transit
planners are often trying to take stops out and/or implement policies
that create consistent and generally wider spacing" (http://
tinyurl.com/yesz6h8). �Weighting the transit score by the number of
stops in the area won't be very useful. �Additionally, the
distance_penalty shouldn't be so great that it significantly reduces
transit scores in the region where the system has widened stop
spacing.

I'm not a fan of the idea of weighting the trip values differently
according to mode � rail vs. bus, for example. �The "distance-
penalized trip-weighted�" approach will show if there is a difference
in service speeds that makes more of the city accessible, anyway. �I
acknowledge that there are differences in ride quality between
different types of vehicles. �But the distinction between bus and rail
in terms of service quality, station amenities, (and who each is for)
etc. should, and is, blurring. �If people really want to be near light
rail, the alignments and stations are usually pretty easy to find even
without a transitscore.

-Aaron
 wrote:


Semantically as it relates to buses and bus stops, I would think one
should clarify that accessiblity is related to (unique) scheduled
trips that serve a bus stop (or stops).

For an address in the "center" of a neighborhood, where the individual
could just as well walk two blocks west to a bus stop bus to catch the
bus - or two blocks north to a different bus stop location to catch
the same bus (the route making an eastbound right turn from one
arterial to the other), the fact of being "served" by two bus stops
only equates to being served by one bus (route) or however many
scheduled trips. �Conversely, the north-south arterial may have only
the branch of one scheduled route direction - while the east-west
arterial could be carrying a trunk of trips from both that route
direction as well as additional trips on another route direction/bus
line.

I'd also note that trips per (typical) week would seem to best
encompass the various daily schedule deviations that might befall a
bus stop location (i.e. weekday, saturday, sunday service, extra trips
late night on a friday only, etc.)

On Nov 4, 3:20�pm, Kieran Huggins <kie...@kieran.ca> wrote:



Since you're looking for a _general_ accessibility score, I think it would
be best to calculate the stoptime frequency graph against a general
ridership volume graph for the city in question. For example: one bus 20
minutes apart during rush hour is *less* valuable than the same bus 30
minutes apart in the middle of the night. Does that make sense?


Might also be sage to take the direction of the service into account as
well, thereby calculating a total "value area" as a combination of vectors,
whose length is proportional to the frequency. Thus, an intersection with
buses in multiple directions would be considerably more valuable, while hubs
and stations would have the highest value. You could also factor the average
speed of each vehicle into that equation (subways > express buses > buses).


Without having more information about the specific visitor (like where they
want to go and when), I'm not sure what else can really be done.


Cheers,
Kieran


On Wed, Nov 4, 2009 at 2:36 PM, fruminator <mich...@frumin.net> wrote:


these are 'isochrones'.


isochones = accessibility only IF you assume the attractiveness of
destinations falls of linearly with travel time. �This is likely not
the case, thus Brandon's question about the shape of the function.


as for Jerry J, the comparison of actual to scheduled running times is
a big deal so I'm glad to hear you are doing that. �Rarely is that
sort of thing aired publicly. �There are a lot of reasons that there
will be discrepancies, some good and some bad. �But to get you
started, let me ask this question -- assume that you had the scheduled
running time and you measure a distribution of running times, then how
would you set the scheduled running time as a function of that
distribution? �the min() ? the max() ? the mean()? the median() ? the
95th percentile? etc. �this gets into a fairly complicated set of
tradeoffs between cost and reliability and passenger wait times and
passenger in-vehicle travel times and layover time at terminals and en-
route holding policies etc.


Mike


On Nov 4, 2:02 pm, Philip Ashlock <p...@openplans.org> wrote:


I'm sure some of you have seen this before, buthttp://


www.triptropnyc.com/


provides a very simple visualization of accessibility based on subway
metrics. Simultaneous correlation with buses would make this even more
accurate.


-phil


On Nov 4, 12:44 pm, Tommy B Goode <tommybgo...@gmail.com> wrote:


For Philadelphia your calculations are very simple at the moment. (:


On Wed, Nov 4, 2009 at 12:40 PM, Jerry J <jjari...@gmail.com>
 wrote:


fruminator gave an excellent response with the concept of
accessibility and those sources. �Many researchers in this field have
translated the metric "how far you can go in 45 minutes" to the


number


of job opportunities or medical facilities (destinations relevant to


a


transit rider) that can be reached in this time.


check out this paper by Robert Cervero: �Tracking accessibility:
employment and housing opportunities in the San Francisco Bay Area
http://envplan.com/abstract.cgi?id=a311259
At Berkeley he is the professor that has done the most research in
this field (transportation and land use)


Also, I haven't read this one but it seems relevant: Measuring
accessibility: an exploration of issues and alternatives
http://www.envplan.com/abstract.cgi?id=a291175


I'm not at school right now, so I don't have access to the PDFs, but
if you are interested in taking a look, I can download them and send
them to you in a couple hours. �Also, maybe a city planner person can
chime in about this also.


Also, I guess I should mention that I am working on something
related. �Transitshed creates travel time maps based on schedule
information. �From GPS traces of buses, I am also creating travel


time


maps, and trying to find areas of major discrepancy between the
schedule and actual performance of the transit system and maybe the
reasons for these problems. �I only have the data now for the sf bay
area (thanks to NextBus) but maybe this can help out with the transit
score in the future?


Jerry


On Nov 3, 12:46 pm, fruminator <mich...@frumin.net> wrote:


Brandon,


What you're basically talking about is well-known as a general


concept


called 'accessibility' which is sort of a reaction to an obsession
with 'mobility.'


These 2 theses should give a decent review and set of references on
the topic of accessibility:


Brandon Martin-Anderson

unread,
Nov 30, 2009, 2:59:00 PM11/30/09
to transit-d...@googlegroups.com
I'm sorry Mike - I was being sarcastic. Also, I forgot to thank you
for the literature links.

My point with the sarcasm is that when a trip is marked "light rail"
or "subway" or some other steel-on-steel fixed right-of-way sort of
transit, that tag counts as a shorthand for "less crowded, more
reliable, fewer stops, smoother ride", which deserves a score boost.
If every GTFS feed came with an amenity table which listed the
crowded-ness, reliability, and ride smoothness of every trip in the
system we could just use that for the service-quality metric and I
would see no reason to preferentially weight rail service.

-B

On Mon, Nov 30, 2009 at 11:51 AM, Michael Frumin <mic...@frumin.net> wrote:
> I can't tell if that's a sarcastic paraphrasing.  I didn't mean to say that
> these studies control for _all_ things.  For example I'm sure they were not
> able to find any steel-wheel-on-rail buses or rubber-tire-on-road trains.
>
> What this is useful for is to help focus attention on what matters --
> protecting transit from private vehicle congestion, improving reliability
> and frequency, etc.
>
> Brandon Martin-Anderson wrote:
>
> I also suspect that if your value analysis controls for all the things
> that distinguish a train from a bus then the value of a train would be
> identical to a bus. :)
>
> On Mon, Nov 30, 2009 at 11:40 AM, Michael Frumin <mic...@frumin.net> wrote:
>
>
> There is infinite empirical research on the subject, and of course none of
> it agrees.
>
> One place to start is with the brits, since they love cost-benefit analysis,
> which depend heavily on relative values-of-time.  eg:
> additional value (as Tim Sobota pointed out).  Maybe this is jumping
> in the weeds sooner than necessary -- I assume this would get worked
> out later.  Jarret Walker (humantransit.org) points out that "transit
> planners are often trying to take stops out and/or implement policies
> that create consistent and generally wider spacing" (http://
> tinyurl.com/yesz6h8).  Weighting the transit score by the number of
> stops in the area won't be very useful.  Additionally, the
> distance_penalty shouldn't be so great that it significantly reduces
> transit scores in the region where the system has widened stop
> spacing.
>
> I'm not a fan of the idea of weighting the trip values differently
> according to mode — rail vs. bus, for example.  The "distance-
> penalized trip-weighted…" approach will show if there is a difference
> in service speeds that makes more of the city accessible, anyway.  I
> acknowledge that there are differences in ride quality between
> different types of vehicles.  But the distinction between bus and rail
> in terms of service quality, station amenities, (and who each is for)
> etc. should, and is, blurring.  If people really want to be near light
> rail, the alignments and stations are usually pretty easy to find even
> without a transitscore.
>
> -Aaron
> http://www.trilliumtransit.com/blog
>
> On Nov 5, 11:36 am, T Sobota <tsob...@cityofmadison.com> wrote:
>
>
> Semantically as it relates to buses and bus stops, I would think one
> should clarify that accessiblity is related to (unique) scheduled
> trips that serve a bus stop (or stops).
>
> For an address in the "center" of a neighborhood, where the individual
> could just as well walk two blocks west to a bus stop bus to catch the
> bus - or two blocks north to a different bus stop location to catch
> the same bus (the route making an eastbound right turn from one
> arterial to the other), the fact of being "served" by two bus stops
> only equates to being served by one bus (route) or however many
> scheduled trips.  Conversely, the north-south arterial may have only
> the branch of one scheduled route direction - while the east-west
> arterial could be carrying a trunk of trips from both that route
> direction as well as additional trips on another route direction/bus
> line.
>
> I'd also note that trips per (typical) week would seem to best
> encompass the various daily schedule deviations that might befall a
> bus stop location (i.e. weekday, saturday, sunday service, extra trips
> late night on a friday only, etc.)
>
> On Nov 4, 3:20 pm, Kieran Huggins <kie...@kieran.ca> wrote:
>
>
>
> Since you're looking for a _general_ accessibility score, I think it would
> be best to calculate the stoptime frequency graph against a general
> ridership volume graph for the city in question. For example: one bus 20
> minutes apart during rush hour is *less* valuable than the same bus 30
> minutes apart in the middle of the night. Does that make sense?
>
>
> Might also be sage to take the direction of the service into account as
> well, thereby calculating a total "value area" as a combination of vectors,
> whose length is proportional to the frequency. Thus, an intersection with
> buses in multiple directions would be considerably more valuable, while hubs
> and stations would have the highest value. You could also factor the average
> speed of each vehicle into that equation (subways > express buses > buses).
>
>
> Without having more information about the specific visitor (like where they
> want to go and when), I'm not sure what else can really be done.
>
>
> Cheers,
> Kieran
>
>
> On Wed, Nov 4, 2009 at 2:36 PM, fruminator <mich...@frumin.net> wrote:
>
>
> these are 'isochrones'.
>
>
> isochones = accessibility only IF you assume the attractiveness of
> destinations falls of linearly with travel time.  This is likely not
> the case, thus Brandon's question about the shape of the function.
>
>
> as for Jerry J, the comparison of actual to scheduled running times is
> a big deal so I'm glad to hear you are doing that.  Rarely is that
> sort of thing aired publicly.  There are a lot of reasons that there
> will be discrepancies, some good and some bad.  But to get you
> started, let me ask this question -- assume that you had the scheduled
> running time and you measure a distribution of running times, then how
> would you set the scheduled running time as a function of that
> distribution?  the min() ? the max() ? the mean()? the median() ? the
> 95th percentile? etc.  this gets into a fairly complicated set of
> tradeoffs between cost and reliability and passenger wait times and
> passenger in-vehicle travel times and layover time at terminals and en-
> route holding policies etc.
>
>
> Mike
>
>
> On Nov 4, 2:02 pm, Philip Ashlock <p...@openplans.org> wrote:
>
>
> I'm sure some of you have seen this before, buthttp://
>
>
> www.triptropnyc.com/
>
>
> provides a very simple visualization of accessibility based on subway
> metrics. Simultaneous correlation with buses would make this even more
> accurate.
>
>
> -phil
>
>
> On Nov 4, 12:44 pm, Tommy B Goode <tommybgo...@gmail.com> wrote:
>
>
> For Philadelphia your calculations are very simple at the moment. (:
>
>
> On Wed, Nov 4, 2009 at 12:40 PM, Jerry J <jjari...@gmail.com> wrote:
>
>
> fruminator gave an excellent response with the concept of
> accessibility and those sources.  Many researchers in this field have
> translated the metric "how far you can go in 45 minutes" to the
>
>
> number
>
>
> of job opportunities or medical facilities (destinations relevant to
>
>
> a
>
>
> transit rider) that can be reached in this time.
>
>
> check out this paper by Robert Cervero:  Tracking accessibility:
> employment and housing opportunities in the San Francisco Bay Area
> http://envplan.com/abstract.cgi?id=a311259
> At Berkeley he is the professor that has done the most research in
> this field (transportation and land use)
>
>
> Also, I haven't read this one but it seems relevant: Measuring
> accessibility: an exploration of issues and alternatives
> http://www.envplan.com/abstract.cgi?id=a291175
>
>
> I'm not at school right now, so I don't have access to the PDFs, but
> if you are interested in taking a look, I can download them and send
> them to you in a couple hours.  Also, maybe a city planner person can
> chime in about this also.
>
>
> Also, I guess I should mention that I am working on something
> related.  Transitshed creates travel time maps based on schedule
> information.  From GPS traces of buses, I am also creating travel
>
>
> time
>
>
> maps, and trying to find areas of major discrepancy between the
> schedule and actual performance of the transit system and maybe the
> reasons for these problems.  I only have the data now for the sf bay

Michael Frumin

unread,
Nov 30, 2009, 3:22:08 PM11/30/09
to transit-d...@googlegroups.com
I agree.  Sort of.

Clearly there are efforts, quite successful in many places, to improve the quality of service of buses (*cough*BRT*cough*).  I would be very hesitant to assume that the service quality of, say, the M14 in Manhattan or the #1 in Boston/Cambridge, is the same as the service quality on the new Viva BRT system in Northern Toronto.  It is too bad that there is no BRT indicator in GTFS (is there?).

All of which sort of gets to my dream of including historically measured performance in trip planners (http://frumin.net/ation/2008/11/whats_in_a_schedule.html)

Mike

John Campbell

unread,
Nov 30, 2009, 3:26:01 PM11/30/09
to transit-d...@googlegroups.com
On Mon, Nov 30, 2009 at 2:59 PM, Brandon Martin-Anderson
<bad...@gmail.com> wrote:
> I'm sorry Mike - I was being sarcastic. Also, I forgot to thank you
> for the literature links.
>
> My point with the sarcasm is that when a trip is marked "light rail"
> or "subway" or some other steel-on-steel fixed right-of-way sort of
> transit, that tag counts as a shorthand for "less crowded, more
> reliable, fewer stops, smoother ride", which deserves a score boost.

My gut tells me the only differences that actually matter are schedule
adherence, and trip length variability. "Less crowded" "smoother
ride" doesn't apply for the subway vs bus scenario (at least in New
York). I assume you would agree that fewer stops is a bad thing if
you are already controlling for travel time.

Of course schedule adherence is not in the GTFS data, but you may be
able to tease out some of the difference between rail and buses by
looking at variability in trip length over the course of a day. It is
probably not worth the effort, but if enough cities have Rail, Bus,
and Bus Rapid Transit, then your model will fail to account for the
fact that BRT is just as good as rail.

Regards,
John Campbell

Kieran Huggins

unread,
Nov 30, 2009, 3:27:19 PM11/30/09
to transit-d...@googlegroups.com
Looking at all the feedback we've had on myttc, I've noticed a strong preference for rail (streetcars & subways) over buses, even when it doesn't make sense from an efficiency point of view. This could be attributed to many factors: schedule reliability, crowding, weather isolation (important during some months Canada!), ride smoothness, etc...

The thing I've come to realize though is that the actual metrics seem to matter less than the end result: people in our region prefer rail. Maybe factors like rail-bias are different from region to region, but can be adjusted based on user feedback?

In short: complete impartiality is a wonderful thing, but it doesn't seem to be what users want from our service. They want the "best" options, in all their subjective, rail-biased glory.

That said, if anyone finds a C implementation of the Happiness algorithm, please post it here ;-)

Jehiah Czebotar

unread,
Nov 30, 2009, 3:35:31 PM11/30/09
to transit-d...@googlegroups.com
On Mon, Nov 30, 2009 at 3:22 PM, Michael Frumin <mic...@frumin.net> wrote:
>  It is too bad that there is no BRT indicator in GTFS (is there?).
>

There have been several proposals on the gtfs-changes mailing list to
increase the detail about route_types. There was a bit of discussion,
but no data providers have jumped on board and tried to implement one
of the proposals (as far as I recall)

feel free to jump over on that list, revive one of those threads, and
mention how it would benefit this use case.

http://groups.google.com/group/gtfs-changes/web/open-proposals

--
Jehiah

Michael Frumin

unread,
Nov 30, 2009, 3:40:08 PM11/30/09
to transit-d...@googlegroups.com
thanks but no thanks.  the last thing I want to get in the middle of, let alone revive, is a debate about the relative merits of bus vs BRT vs light rail vs heavy rail etc, as if those are even well-defined terms.  My hope here was just to provide some links to empirical evidence that people were asking for.

my longer-term preference would be for providers to spit out historical performance data in standard formats so that we could work on ways to incorporate that performance into trip planners, so if that debate is raging anywhere please let me know.

thanks,
Mike


Jehiah Czebotar wrote:

--
Jehiah

--

You received this message because you are subscribed to the Google Groups "Transit Developers" group.
To post to this group, send email to transit-d...@googlegroups.com.
To unsubscribe from this group, send email to transit-develop...@googlegroups.com.
For more options, visit this group at http://groups.google.com/group/transit-developers?hl=en.


  

Jehiah Czebotar

unread,
Nov 30, 2009, 3:48:30 PM11/30/09
to transit-d...@googlegroups.com
the discussions on gtfs-changes isn't about the merits of each of
those, just the proper way to represent and describe them in GTFS,
which is quite different.

I will agree though that it would be good for a standard to emerge
around historical performance data, and for it to become common for
agencies to publish that data.

--
Jehiah

Brandon Martin-Anderson

unread,
Nov 30, 2009, 4:01:51 PM11/30/09
to transit-d...@googlegroups.com
Maybe define transit service classes as a set of description tags.
"subway:[level:below ground, surface:steel rail, right of way:
exclusive]" etc.

-B

Aaron Antrim

unread,
Nov 30, 2009, 4:08:01 PM11/30/09
to transit-d...@googlegroups.com
I agree that rail travel modes are more often reliable (less affected by traffic), faster (wider station spacing), and offer a more comfortable ride. These advantages (other than ride quality, which is harder to emulate with rubber-tire vehicles) are often achieved through features and amenities that are packaged as part of rail, but don't necessarily have to be.

Bus Rapid Transit projects can achieve rail-like reliability with signal priority and dedicated lanes or separated travel ways. The more advanced rubber-tired vehicles used in these projects offer at-grade boarding, higher ceilings for larger, airier interiors, and better ride quality than the average bus (but not equalling rail). BRT projects can also involve spacing out stations. The stations are generally nicer, more "rail"-like, with amenities like upcoming arrival screens.

I like and recommend this post (and the comments) on the bus vs. rail issue quite a lot: "Streetcars: an inconvenient truth" (http://www.humantransit.org/2009/07/streetcars-an-inconvenient-truth.html). And, subsequently, "mobility vs. access" (http://www.humantransit.org/2009/07/what-i-meant-by-mobility-could-also-be-called-access.html).

The issues I see with weighting the value of trips differently by mode are:
(1.) Such a simple approach will mislead in cases where regions have invested in amenity-rich or rapid bus services. Is rail *always* superior to bus?
(2.) Ride quality may not be important to every user, but access probably is.

Deciding bus vs. rail becomes a political issue for some communities. Walkscore's goal is to help people to value walkability features. Biasing for rail in transitscore may further reinforce people's association of rail with all things good in transit, and bus with all things bad (I am being a bit dramatic/overstating).

The overall problem I see that's inspired me to bring this up is that transit planners can't always use the most appropriate/cost-effective means of delivering service right now because citizens/feds, etc. have all these historical notions about what what rail and bus services can and cannot be and who they are for. That leads to cost inefficiency and poorer decision-making. Also, some passengers end up not seeing a "transit system" so much separated "rail systems" and "bus systems."

I think it makes sense for the transitscore to be a measure of access. If rider experience or amenity quality is considered, this could be more accurately assigned using information from a source like the proposed stop_features.txt (http://groups.google.com/group/gtfs-changes/web/stop-amenity-proposal).

In any case, I'd like to suggest it is important that any transitscore rail/bus weighting scheme is fairly transparent to users.

Transit mode debates are fun :)

-Aaron

Brandon Martin-Anderson

unread,
Nov 30, 2009, 4:34:05 PM11/30/09
to transit-d...@googlegroups.com
Thanks Aaron. That's a lot of really great feedback.

-B

Aaron Antrim

unread,
Dec 1, 2009, 1:58:18 AM12/1/09
to transit-d...@googlegroups.com
Glad it was useful.

After I blogged a little on the transitscore effort (http://www.trilliumtransit.com/blog/2009/11/19/whats-your-transit-score/), Jarret Walker got in touch with me with some feedback on how to develop a transit score. I told him I'd pass that along here…

> Walkscore already has a map that shows where you can get to on transit in a fixed amount of time from a given address, right? The one I linked to here:
>
> http://www.humantransit.org/2009/07/what-i-mean-by-mobility.html
>
> I assume that this map is based on Google Transit data. I assume that they get total travel time by summing walk time, in-vehicle travel time and half the frequency for the waiting time. Is that right? Because that's how we calculate travel time in the real world.
>
> That really covers everything you need to know to determine if transit's useful. So why couldn't you do a transit score based simply on the outputs of that analysis that's already been done. All you need to do is compress the map into a number, and you could do that by looking at what percent of the city's population and jobs are in that area, or something like that.
>
> Isn't that the way to calculate a true transit score, one that really tells you how much transit access you'll have from a particular point?

Roger Slevin

unread,
Dec 1, 2009, 2:45:00 AM12/1/09
to transit-d...@googlegroups.com
I have been lurking on this topic - but I am now prompted to make a
contribution from the UK where a lot of work has been done on "accessibility
modelling" for transport planning purposes - trying to evaluate how
accessible various locations are in overall terms. If you just add the
component times together all you get is a measure of how quick it is to get
from A to B. What's missing are
- perceived values of walk time and wait time compared with in-transit time
... typically people weight walk and wait times more highly than in-transit
times
- what range of destinations are accessible as a result of the service
pattern available (and if you are more sophisticated you would break this
down into different activities at those destinations - so accessibility to
employment opportunities, shopping opportunities, healthcare opportunities,
educational opportunities, etc)
- how frequency is perceived and used - for high frequency services there
was work in London many years ago that showed that average wait time is
equal to (and not half) frequency in such situations because the
perturbation from schedule was high ... whilst once you are down to
frequencies of perhaps 15 mins or greater then the perceived "delay"
implicit in frequency is not half-frequency, but a lower value as travellers
adjust to known availability of services

I don't think there can ever be one score that is valid for all purposes -
but there are different approaches which satisfy different requirements.
Some might be very simplistic - but others could be much more complicated.
If these scores are to mean anything in terms of planning policies then I
would favour the more complex approaches which reflect human perceptions
rather than absolute values. If you are trying to get a simple indicator to
be able to show area A is more public transport accessible than area B,
maybe the simpler approaches are fine.

Roger

-----Original Message-----
From: Aaron Antrim [mailto:aa...@arcatacommunity.org]
Sent: 01 December 2009 6:58 AM
To: transit-d...@googlegroups.com
Subject: Re: [transit-developers] Re: Developing a "Transit Score" for all
locations

Glad it was useful.

After I blogged a little on the transitscore effort
(http://www.trilliumtransit.com/blog/2009/11/19/whats-your-transit-score/),
Jarret Walker got in touch with me with some feedback on how to develop a
transit score. I told him I'd pass that along here.

> Walkscore already has a map that shows where you can get to on transit in
a fixed amount of time from a given address, right? The one I linked to
here:
>
> http://www.humantransit.org/2009/07/what-i-mean-by-mobility.html
>
> I assume that this map is based on Google Transit data. I assume that
they get total travel time by summing walk time, in-vehicle travel time and
half the frequency for the waiting time. Is that right? Because that's how
we calculate travel time in the real world.
>
> That really covers everything you need to know to determine if transit's
useful. So why couldn't you do a transit score based simply on the outputs
of that analysis that's already been done. All you need to do is compress
the map into a number, and you could do that by looking at what percent of
the city's population and jobs are in that area, or something like that.
>
> Isn't that the way to calculate a true transit score, one that really
tells you how much transit access you'll have from a particular point?



On 30 Nov 2009, at 1:34 PM, Brandon Martin-Anderson wrote:

> Thanks Aaron. That's a lot of really great feedback.
>
> -B

Brandon Martin-Anderson

unread,
Dec 1, 2009, 2:05:34 PM12/1/09
to transit-d...@googlegroups.com
Regarding the question "are fixed-infrastructure modes more valuable
than rubber-wheel modes?" Jarrett Walker sums up his position nicely:

"My observation in the streetcar post was that if you replace a bus
with a streetcar, and make no other changes, nobody can get anywhere
any faster than they can now. In other words, such a project does
nothing to expand the area of access shown by this tool, for anyone."

The prop that Jarrett is using for this argument is something that I
made, and so I feel well-qualified to comment on some of its
philosophical weaknesses.

On Mobility vs. Accessibility: I am much more interested in increasing
the amount of cool stuff at my disposal, not the shed of dirt I can
walk on in a fixed interval of time. There is no use in having great
transit access to a desert. So, maybe the next town over has a better
gym than my town. The solution to that problem is to either get a
better gym in my town or move to that town - not to spread the whole
of human artifice in a thin accessible sheet over the surface of the
earth. The very _definition_ of a city is a spatial concentration of
amenities; the driving principle of their formation is the
desirability of that concentration.

On the relative desirability of street cars: the isochron map uses
time as the trip weighting factor because it's easy to understand, but
in reality trips are not planned to minimize time - they are planned
to minimize generalized cost as a function of all the combined
inconveniences of travel. A more "true" transit time map would not map
isochronic contours - it would map isoeconomic contours - a boundary
around everywhere someone can get _in a given amount of
inconvenience_. This is the way people work in reality - their
attachment to travel time is strangely tenuous. They hate transferring
buses, or walking more than a few blocks, even when it saves them a
significant amount of time. Anything you do to bring down the
generalized cost of a trip without necessarily bringing down the
travel time _will_ increase accessibility (and mobility) because it
will act to extend the boundary of all amenities reachable within a
certain amount of inconvenience. Street-car riding has a lower
generalized cost per unit time than riding on a bus, so replacing a
bus with a streetcar will increase the accessibility of those who use
that service.

-B

Aaron Antrim

unread,
Dec 1, 2009, 8:46:58 PM12/1/09
to transit-d...@googlegroups.com
Yes.  There are a lot of factors that influence travel choice.  Many to most WalkScore users will prefer rail.  However, weighting by mode will mean making some generalizations about modes that may be true most but not all of the time.  These assumptions also frustrate the effort to de-package modes with service features and amenities.

DHofmann

unread,
Dec 3, 2009, 8:42:28 PM12/3/09
to Transit Developers
On Nov 30, 10:58 pm, Aaron Antrim <aa...@arcatacommunity.org> wrote:
> > Isn't that the way to calculate a true transit score, one that really tells you how much transit access you'll have from a particular point?

No, transit access is only a means to an end. A true transit score is
one that tells you how convenient it is to get to actual places.
Convenience means:
1. Time from doorstep to destination. Less is better.
1a. Distance to a destination is irrelevant--a store 10 minutes from
home by bus is the same to me as one that is 10 minutes by bullet
train, even though one is much farther than the other.
1b. Frequency is also irrelevant--waiting 20 minutes for a fast bus
that then gets me there in 10 minutes is the same as waiting 10
minutes for a slow bus that then gets me there in 20 minutes.
1c. Doorstep to destination should include any time spent walking from
doorstep to bus stop, time waiting for the bus and layovers, and time
spent walking from bus stop to destination.
2. Cost. Less is better.
3. Operating hours. More is better. Peak transit periods are more
important than off-peak.

Aaron Antrim

unread,
Dec 4, 2009, 1:55:27 PM12/4/09
to transit-d...@googlegroups.com
DHofmann:

I agree with you, mostly.  But your metrics seem to me to be about access (and not amenities nor ride quality).  It's a matter of semantics, I suppose.  Speed of service, walking time, etc. all do affect access.

Is frequency really "irrelevant"?  It seems highly relevant -- as you say high frequency service can reduce waiting time, and therefore overall travel time.  Note higher frequency often distinguishes rail from bus service -- but there are high-frequency bus lines too.

DHofmann

unread,
Dec 4, 2009, 7:19:16 PM12/4/09
to Transit Developers
Aaron, frequency of service is only relevant when you don't add the
average time waiting for the bus into the total transit time. I
suppose this is semantics again.

As for ride quality, I would rather ride a fast bus (BRT) than a slow
train. To me, ride quality is irrelevant when total transit time is a
metric.

Brandon's post (which I hadn't seen until after I posted) brings up a
good point about transfers being undesirable, even when it saves time.
So I'd like to add:

4. Complexity of itinerary. Less complex (fewer transfers) is better.

Derek

Edward Vielmetti

unread,
Dec 4, 2009, 9:22:40 PM12/4/09
to transit-d...@googlegroups.com
The biggest transit metric I face daily is the predictability
of delays, either on a regular basis or at a specific time.

For instance, I know that I can count on taking the 5:03 bus
unless it is more than 17 minutes late when it arrives at my
destination because that gives me enough time to get to
day care on time; but I can't take the 5:18 because the buffer
is only 2 minutes and it's that late too often. And I know from
bitter experience that if there's so much snow that the buses
are running 20+ minutes late that I need to take the 4:48 or
be late (or perhaps better yet just walk, because at that point
I can walk faster than the bus).

As another instance, it's much more pleasant to wait for
an every 15 minutes bus if you know that it's running
6 minutes late this one time, because that's 6 more minutes
you can spend in the warmth of your home or cafe or office
before running to catch it.

Some dreamed up route analysis software that I am sure
does not exist takes a standard printed timetable and does
a color-coded overlay to note for each scheduled time stop
what the likelihood of an on-time arrival is and what the expected
delay is, similar to airplane on-time arrival statistics. I'd note
that routes that sometimes run late, sometimes run early,
and that vary widely for the same route at the same time of
day are the worst, because you can never know if you missed
your bus because it was late, early, or that was the previous bus
running 13 minutes late that went by.

Ed

Edward Vielmetti
AnnArbor.com

"i'm already 15 minutes early for the next bus"
> --
>
> You received this message because you are subscribed to the Google Groups "Transit Developers" group.
> To post to this group, send email to transit-d...@googlegroups.com.
> To unsubscribe from this group, send email to transit-develop...@googlegroups.com.
> For more options, visit this group at http://groups.google.com/group/transit-developers?hl=en.
>
>
>



--
Edward Vielmetti
Ann Arbor, MI 48104

Google Voice: +1 734 330 2465
Web: http://vielmetti.typepad.com

jed horne

unread,
Dec 21, 2009, 10:51:24 AM12/21/09
to transit-d...@googlegroups.com, Jarrett Walker
You'll also run into different issues with headway - people tend to
weigh waiting time somewhere from 1.5 to 2 times worse than riding,
and long headways become a problem if you assume some x percentage
variation in when the bus arrives times the consequence of missing it.
For an individual who wants a real-time statistic on current
accessibility this isn't so bad since you're not dealing with
stochastic arrivals, but aggregating over the course of the day or
over a population is problematic.

Also, this is kind of neat for visualizing isochromes:

http://www.tom-carden.co.uk/p5/tube_map_travel_times/applet/

Sorry for waiting so long to respond to this,

Jed

Reply all
Reply to author
Forward
0 new messages