(SR-52) Simplified Methods in Transportation Analysis

[SlideRule]

A excerpt from SIMPLIFIED METHODS IN TRANSPORTATION ANALYSIS, Serge Tsygalnitzky, Master of Science - Massachusetts Institute of Technology, May 1977, 126 pages

"                                        ABSTRACT
           Submitted to the Department of Civil Engineering
            on May 11, 1977 in partial fulfillment of the
           requirements for the degree of Master of Science

This thesis is concerned with the development of simplified methods of analysis in transportation systems. The objective is to find ways to predict the impacts and estimate a demand model, which are not based on large scale data or computer-based analyses. Instead, the methods are designed to be used on programmable pocket calculators for ease of analyst use.
…
The methods are intended to be useful to analyse decisions from the perspective of a transportation operator. Among the-specific methods developed, the most important are:

     - Design of a simplified way of forecasting using a logit model;
     - Estimation of a logit model with least squares method;
     - Inferences about the context when a lot of data is missing, as
       obtaining of O-D matrix for all modes when the characteristics
       of only one mode are known.
 In conclusion, directions for future work are also indicated. The results of this thesis demonstrate that useful simplified methods can be developed, and thus open up an important new field of research.
…
     We will also develop the model estimation giving some examples for the two basic cases: modal split model and multiple choice model. We will then study a simplified way of running multiple linear regressions in order to be able to program them on a pocket calculator.
…
     Lastly, we will develop some methods that are within the capacity of our pocket calculator. (Several programs and worksheets are included in the appendix.) As time passes the capacity of these pocket calculators increases, and these methods can very well be outdated in a few years. Also, it is important to remember that many of these methods are empirical and have no mathematical argument. Their only quality is that they give fair results according to few experiments.
…
… demand model should have two qualities: it should be estimable by hand or pocket calculator, and readily applicable for a forecast.
…
… the calibration of aggregate models is often based on linear regression (that can be programmed on a pocket calculator) while the disaggregate ones are based on maximum likelihood because the data are discrete (the data are based on a binary Yes/No answer from some surveys)
…
     As we are looking for a simplified method, this method seems more suitable for our concern, because it can be programmed on a pocket calculator. Therefore we will use it in the rest of the study.
…
     We must not forget that all the methods were developed because they could be programmed on a pocket calculator and did not need any computer hardware.
…
SR-52 User Instructions
     LOGIT MODEL BASE CASE
     LOGIT MODEL PIVOT POINT
     LOGIT MODEL ADDITION OF A NEW ALTERNATIVE
     O-D MATRIX
     LINEAR REGRESSION "
with program listings

BEST!
SlideRule