Team E

100/100 for both I3 and I4.

Tian Wu, Tim Higgins, Matt Belenky, Chao Wang, John Sperger

General Comments:
----------------

This project will be great.  However I want to make some fairly
specific suggestions about your sampling frame/sampling plan.  See
below.


Specific Comments:
-----------------

* The list of previous research in I3 is fine, but it is not formatted
  in a way that is attractive or easy to read in a list of references
  at the end of a paper (and these need to be in your final paper).
  Please reformat.

* SAMPLING PLAN:

  I think a simple random sample of 1/2-hr observation periods over a
  week will obscure some important time of day differences.  So I'd
  like to suggest that you begin by stratifying into four 3-hour
  periods each day:

  7am--10am,
  10am--1pm, 
  1pm--3pm, and 
  3pm--7pm

  Within each stratum, over a 2-week period, there are 30 one-hour
  periods and 60 1/2 hr periods.  You can decide whether you want to
  sample 1/2-hour or 1-hour periods in each stratum.  

  I think 1/2-hour may be too short: fewer buses can be observed in
  1/2 hr than 1 hr, but roughly the same number will be late past the
  end of a period (whether it is 1/2 hr or 1 hr) and hence unobserved.
  This is like a "nonresponse" problem, and the easiest way to fix it
  is to take longer periods (e.g. 1 hr vs 1/2 hr) so as to make the
  nonresponse a smaller fraction of the total number of buses intended
  to be observed in each time period.

  Thus you will have a certain number of (1/2-hr or 1 hr) periods in
  each "stratum", spread over a 2-week period.  I suggest you do
  observations in pairs, one person stationed at the inbound bus stop
  and one at the outbound bus stop, at Forbes and Morewood.  (If you
  think bus punctuality will be different for inbound and outbound
  buses, you might consider 8 strata instead of 4: the 4 above for
  inbound, and for outbound, buses).

  For a first sample size calculation for II.5, pretend you are doing
  a simple random sample and imagine trying to estimate the mean
  number of minutes early or late the buses is.  A graph of the
  distribution of this variable can be found in "Chart 3" from
  Macaffe.  It looks to me like the distribution is log-normal or
  Erlang or something like that.  You can make an eyeball guess of the
  mean (bigger than the mode!) and SD (be generous, since there is a
  long right tail), and use these in the sample size calculation.

  This sample size will be the number of *buses* that you will need to
  observe, and you can convert this into a number of 1-hr (or 1/2-hr)
  periods that you will need, by considering the (average?) number of
  buses per period (less "nonresponse" at the end of each period).
  Note that the number of buses per observation period will be
  different in the different strata above.  

  For more careful calculations later in the project, I think you want
  to think of each observation period as a cluster, in which you
  observe all the units (all the buses).  

* The other things you are planning to observe seem fine.  Please see
  both papers below, but especially Farid (2010) for some other
  measures that might be interesting (not sure whether you can collect
  this data though).

* Please read the following on-line, as examples of other bus
  punctuality/service surveys.  

  - Macaffe, K. (2008).  Bus Punctuality Statistics GB: 
    2007.  Great Britain: Department for Transport.  

    The link to the pdf is in the upper right corner of the page at

    http://www.dft.gov.uk/pgr/statistics/datatablespublications/public/buspunctuality/buspunctuality07

    [this is related to one of your references; it is a paper that is
    well worth reading for your project!  See especially the Annexes
    (appendices) at the end) for ideas about how to record and analyze
    bus punctuality.  Note that it is basically easier to measure
    punctuality for departure times than for arrival times.  It may
    also be useful to compute the excess waiting time using the ideas
    in Annex 2.]

  - Farid, A. (2010).  Trip productivity evaluation of bus service:
    Medan Kidd Bus Station.  Universiti Teknologi Petronas: Presented
    at the Malaysian Universities Transportation Research Forum and
    Conference (MUTRFC 2010).  

    The pdf is at 

    http://www.uniten.edu.my/newhome/uploaded/coe_civil/mutrfc2010/001%20PG%20presentation.pdf

    [this considers a number of different measures in addition to bus
    punctuality]

===