The basic assignment

Each group is to develop statistical models that predict the traffic generated by fast-food restaurants:  one (or more) for "lunch-time" (11:30 -1:30) and one (or more) for "dinner-time/PM peak period" (4:00-6:00).  The structure of the different models will be compared to one another as will the predictions that are made by using them.  The models that you develop and their predictions will also be compared to those from software promulgated by the Institute of Transportation Engineers.  Each group must submit its findings in a comprehensive written report.

Critical dates for project #1

Friday, 30 January:  the outline of the data collection plan is due no later than the beginning of class—it is strongly suggested that you submit this earlier than the due date.

Friday, 6 February:  the data for your site are due no later than the beginning of class (earlier would be better).
 
Monday, 2 March:   the actual project is due no later than the beginning of class.  Please note that on this date you will participate in an exercise during which you will assign grades to your project, your own performance, and the performance of other members of your group.

Some background

The classic models (starting in the late 1950s) for trip generation were linear regression models based on aggregate zonal data.  These types of models are still developed and used today.  Moreover, such models form the basis for most software on the market, including the widely-used Institute of Transportation Engineers (ITE) trip generation software.  As will be noted in class, the development and use of these models is not without some problems.  They include:  1) independence of trip generation estimates from system characteristics; 2) exclusion of appropriate variables; 3) inclusion and use of counter-intuitive or illogical variables (or signs); and 4) lack of temporal consistency of models.

The most widely used models for trip generation for specific sites are found in the ITE trip generation software.  In these models, the trips generated (produced by and/or attracted to a site) are generally functions of the size of the development and the specific land use (e.g., gross floor area [GFA] for restaurants) and/or other measures of the intensity of the use (e.g., number of seats for a restaurant, number of beds for a hospital, number of employees for offices).  These models are typically univariate regression models, an example of which is shown below:

where:   Yt  = total number of daily trips generated
             X1 = gross floor area
             a,b = coefficients determined through regression analysis

However, at least some of the models for fast-food restaurants (based on GFA, number of seats) have been widely criticized for giving inaccurate results.  Thus, there is need for improvement of these models.  (That's what you are going to do!)

Specifics of the assignment

Site selection

Each group will be responsible for choosing one fast-food restaurant and collecting data on its operations and characteristics.  Those data will be shared with all other groups.  The restaurant that you choose must offer both drive-up and "sit-down" service and provide off-street parking.  The target restaurants are the large chain-type places like McDonald's, Burger King, and Taco Bell.  All groups must select a different restaurant (while, for example, more than one McDonald's may be selected, no two groups can select the same McDonald's).  You may make your selections as soon as possible—it’s first-come, first-serve.

Data collection and presentation

A not insignificant part of this project is data collection.  You must collect data on the size and other physical characteristics of the restaurant, the traffic entering and leaving the site, and the traffic on the adjacent roadway(s).

You will need to collect the following traffic data organized in 15-minute intervals:

     * traffic volume entering the facility (VOLenter)
     * traffic volume leaving the facility (VOLexit)
     * number of vehicles using the drive-up window(s) (drvupVOL)
     * maximum queue length at the drive-up window(s) maxqueue)
     * traffic volume on the adjacent roadways (each direction, separately, for any major
       roadway [nearVOL, farVOL for nearest and farthest lane from facility,
       respectively]; combined for minor roadways [minorVOL])

You will also need to collect the following data about the restaurant itself:

     * gross floor area (GFA)
     * number of seats (seats)
     * number of employees on site during traffic data collection periods (employee)
     * number of parking spaces (park)
     * number of actual drive-up windows where food is ordered (driveup)
     * number of registers/stations excluding the drive-up windows (register)

Each group will have to collect data on at least three separate (one hour) occasions:  one or two "lunch-time" periods and one or two "dinner-time" periods (two of one, one of the other).  A total of three hours (12 15-minute periods) of data is required from each group.

Once the data are collected they must be put into a plain ASCII file (e.g., as made by a DOS text editor or PC-Write) and turned in.  The precise format for the data is provided in a separate handout—YOU MUST USE THE FORMAT SHOWN (EXACTLY).  Your data will be combined with data from all other groups and returned to you for model calibration.  Errors will result in a lower final grade.

All data must be collected and turned in (both a paper copy and a diskette containing one file containing the data) no later than Friday, 6 February, at the beginning of class.  If you do not turn your data in on time, your final project grade will be penalized one full letter grade.

Model calibration

You will need to calibrate several simple linear regression models which relate the number of trips to and from fast-food restaurants to characteristics of the restaurant and/or traffic volume information.  You will use the data that were collected by all groups (past and present) to calibrate the models discussed below.

Traditional ITE-type models

The traditional ITE models relate total driveway volumes to characteristics of the restaurant.  You will create ITE-type linear regression models using the dependent and independent variables listed below (a total of 24 simple univariate models).

   dependent variables (see NOTE below)

   TLUNCH:   total hourly driveway volume during lunch period
   TPMPEAK:  total hourly driveway volume during PM peak period
   INLUNCH:  inbound hourly driveway volume during lunch period
   INPMPEAK:  inbound hourly driveway volume during PM peak period
   WLUNCH:   total hourly service window volume during lunch period
   WPMPEAK:  total hourly service window volume during PM peak period

   ITE-type independent variables

   GFA:   gross floor area
   SEATS:  number of seats
   EMPLY:  number of employees on site during traffic data collection period
   PARK:  number of parking spaces

For each of the dependent variables listed above, it is necessary that you select the model which is "best" (i.e., which regression model best predicts each dependent variable).

448-type models

Since it has been argued by various traffic engineers that, for example, volumes on adjacent roadways may be better predictors of fast-food restaurant driveway volumes than the restaurant characteristics, you will also develop models based on that type of independent variable.  For the 448-type models, the dependent variables will be exactly the same as those listed for the ITE-type models.  However, the independent variables will include variables such as:

   * directional (or total) traffic volumes on adjacent major roadways
   * 2-way volumes on adjacent minor roadways

and possibly be used in combination with one or more ITE-type independent variables.  That
is, the independent variables might be different and the models might be more complex:

where:   Yt   =  total driveway volume for lunch period
             X1  =  restaurant characteristics (e.g., gross floor area)
             X2  =  traffic volume measure for adjacent street/road
       a,b1,b2   = coefficients determined through regression analysis

There will be six 448-type models, one for each dependent variable.  These will be the "best" that can be produced using the data available.

NOTE: The models will be calibrated using data based on 15-minute intervals while the desired estimates are for hours.  Thus, the predictions that the models provide will have to be multiplied by four to get hourly volume estimates.

Comparisons of Trip Estimates

Once the ITE-type and 448-type models are calibrated, they will be used to make various predictions (estimates) of trips generated by your site. These predictions will be compared to those obtained by using the ITE trip generation software and the actual field observations that you made.

Thus, in addition to making estimates using the models that you developed, it will also be necessary to use the ITE trip generation software (available on transportation laboratory computers and/or to each group) to predict the various driveway volumes.

As noted above, once the models are calibrated, you will use them to make predictions for all dependent variables for your group's site.  You will need to complete a table similar to the one shown below and thoroughly discuss the differences in the predictions between the various models
and the values observed in the field (note that it is possible that not all cells can be filled).
 
 

A final report must be prepared which includes:  a description of what you did; the results of your analysis including the models that you developed; discussion of the quality of your models and the differences between the predictions of the models for your site and the observed values; and overall conclusions regarding using the various methods and models to predict the traffic related to fast-food restaurants.

Questions/comments about the project

I will be glad to answer questions about the project in class, over e-mail, or during office hours.  If questions are asked during office hours, you may be requested to "ask it again" in class or on e-mail in the interest of maintaining a "level playing field" for all of the groups.

You will need to use a statistical package (e.g., the Statistical Package for the Social Sciences [SPSS]) to do the actual modeling.  A TA will (hopefully) also have "hours" to help you with SPSS—these will be announced in class.  SPSS is available on the machines in the transportation laboratory.  There is also a separate hand-out about using SPSS on the transportation laboratory computers.  (It is not a requirement that SPSS be used—you could do this using EXCEL or another statistical package.)

Outline of data collection plan

For this project, the outline (due no later than Friday, 30 January, at the beginning of class) must cover the data collection plan for your group.  This should include the following points:

1. Identification of your project site—approval may be obtained prior to the submission of the
    outline (prior approval is encouraged as site are assigned on a first-come, first-serve basis).

2. A time schedule for collection of the required data (no weekends, weekdays only).
 
3. A description of how the data will be collected.  This should be based on a field visit that
    included collecting some preliminary data so that it is clear, for example, where data collectors
    will have to be situated so that they can, in fact, collect the required data.  This would include a
    copy of all forms to be used during data collection.

4.  A description of how data will be transferred to a computer disk (e.g., what software will be
     used).

Comments on the final project report

1. The report must "stand alone" and contain a presentation and discussion of the assignment, the
    analysis, the results, and the conclusions—as if you were consultants reporting to a city agency.

2. The report does not have to contain an extensive review of regression procedures, but those
    aspects of the approach which are germane to the analysis must be included.  For example, it
    would be appropriate to discuss some of the assumptions that are inherent in the methods used;
    the effects of violating these assumptions; and the effects of making the assumptions in the first
    place.

3. The report must be of "professional caliber."  (See the "information and some policies on..."
    handout for additional details.)

4. There are several identifiable and separable parts to this assignment.  You must identify (in an
    appendix or acknowledgement) which group members did which parts of the assignment
    individually and which parts were the result of collaboration.  Each student will receive a grade
    for their "part" and the project as a whole.  Any group can also opt for only one collective
    grade.  The decision regarding separate or collective grades must be indicated either in a preface
    or appendix to the report.

5. The report is due on Monday, 2 March 98, at the beginning of class.

6. Approximately 10% of the total grade for this project is in "presentation points."  That is, a
    technically perfect project that is sloppily prepared and/or presented could result in a maximum
    "score" of 90%.  (Trust me on this—the 10% will be taken off.)

Some additional comments on the overall assignment

People often ask me what do I "really" want when I give out an assignment.  This is a fair question.  And the answer is...I want a product that is:  technically correct; well developed; thoroughly explained; well organized; and clearly presented.  If assumptions are made, state them; if decisions are made, explain and defend them; if references are used, cite them.  Some hints on the analysis that is expected (this list is not necessarily exhaustive):

1. The data should be inspected both statistically and graphically for correlations and the general
    relationships between key variables.

2. Alternative forms of the variables should be examined (maybe they work better)—for example,
    what is the impact of using variable X and, alternatively, log(X) or X2 on the predictive
    capability of a model.

3. Your group must decide (and support the decision) whether the "best" model includes all
    possible explanatory variables or only some specified subset of them.

4. The model should also be evaluated from a pragmatic standpoint—does it make sense, is it
    realistic?  You are supposed to be modeling causality—does your model reflect this?

5. While not every model examined or every plot of the data that is made has to be (or should be)
    shown in the report, representative examples should be shown and discussed.   The final models
    must be presented in the body of the report. This project is meant to be done independently by
    group—that is, each group should work separately.  It is, however, permissible to talk about
    general concepts, problems with the analysis package, and other general topics among your non-
    group peers.

For a design project there are generally not "right" and "wrong" answers per se.  There are, however, degrees of achievement.  While attaching absolute grades to your submissions is difficult, it will be done.  If there are significant differences in the levels of achievement by the different groups, there will be similar differences in the grades received.  To that extent there is competition among the groups.  If one group's project clearly stands alone as the best one, they will get an "A"—by comparison, other groups will get less than an "A."  Universal mediocrity will, however, be rewarded with universally mediocre grades.  Each group must indicate which group members are specifically responsible for which part(s) of the project.

On the day that the project is due, you will be participating in a “self-grading” exercise—see following page.

 CE 448
Transportation Planning (Spring 1998)
Sample student self-grading for project 1

A form such as follows will be filled out in class on Monday, 2 March 98, the day the final report on project 1 is due.
--------------------------------------------------------------------------------------------------------------------
You now get to help grade the projects.  Please fill out the matrix below to the best of your ability, be honest and fair.  All results are confidential--they are, however, taken into account during grade determination (both for the project and at the end of the term).

Fill in each group member's name across the top (yours too).

Assign one number for each column for each activity according to a 1 to 10 scale where 1 is the lowest and 10 is the highest.  A "1" indicates that the person did nothing (and something was expected) or that the quality of whatever was done was very low.  If a person did little or nothing for an activity, but that was in accordance with the way the tasks were assigned, insert an asterisk (*) in the appropriate column or leave blank.
 

What overall grade do you think the project should get?_______
 

Any other comments?
 

 CE 448                                                                                                             group#:___
Transportation Planning (Spring 1998)                                                   group grade:___
Project 1—instructor’s grade sheet

Is there an appropriate introduction to the entire project?

Is the site identified and described; graphic?

Is the data collection procedure discussed?

Is the model calibration procedure adequately described?

Is it clear how models were "selected" (e.g., statistical tests)?  Is discussion correct?

Are 24 traditional ITE-type models  developed, shown, and discussed?

Are 6 448-type models developed, shown, and discussed?

Are standard ITE software estimates provided?

Are the estimates of models appropriately compared?

 

		Estimated trips/volume	Estimated trips/volume	Estimated trips/volume
dependent variables	observedtrips/volume	ITE software	ITE-typemodels	448-typemodels
TLUNCH
TPMPEAK
INLUNCH
INPMPEAK
WLUNCH
WPMPEAK

 
Is there an appropriate discussion of the comparison of estimates?

Are there overall conclusions regarding the use of various methods and models?

Is the final report professionally prepared, well-written, and well-organized.

What portion of the 10% of grade does group get for "presentation points?"

Group/individual grade indication?

Other comments: