This page is open for editing because it is part of the Incubator. Have something to add? Please register so you can contribute. Have an option you would like to share? Please click on the 'Talk' button to enter the dialogue. The TF Resource Volunteers appreciate your feedback and interest.

Go to the Project-level traffic forecasting topic page.

Guidelines for Performance Measures

MOEs are often more aggregated than individual traffic volumes, so percent of random error in any MOE may be less than the percent of random error in a single traffic volume forecast. However, the amount of error in any MOE is difficult to estimate, so there is no expectation that error ranges be reported for MOEs.

Reporting requirements differ across MOEs. At a minimum, the analyst should report the name and version of any software product, the nature of any travel forecasting inputs, the source of any parameters, a description of any assumptions, and a succinct statement of the MOEs, preferably in tabular form.

Guidelines for Refinement

The nature of all refinement steps should be indicated. Citations should be made to these guidelines or to NCHRP Report 765, where appropriate. Unrefined forecast data need not be reported. Validation tests for unrefined forecast data need not be reported, so long as the unrefined forecast data meets minimum quality standards as stated in these guidelines.

Factors developed specifically to support a refinement, such as time-of-day factors and directional split factors, should be reported.

The sources of all factors that had be transferred or adapted from elsewhere should be reported. The sources of procedures and parameters should be reported. Citations made to “Trip Generation”, when appropriate, should indicate the edition, land-use, and independent variable.

Results of any refinement that supports project decision making should be reported. These results potentially include refined volumes, refined turning movements and refined speeds. Volumes should be reported directionally, not bidirectionally, for the peak hour or design hour. Tabular presentation of forecast results is preferred to data embedded within paragraphs.

If results from a travel forecasting model are being used, without a refinement step, then reference should be made to any previous validation tests for this model. The base year and any forecast years for the original travel model must be indicated. Any interpolation between years should be described. Any adjustments due to select link analysis should be described.

Guidelines for Windowing and Subarea Focusing

This section describes special reporting requirements for project forecasts that involve windowing or subarea focusing.

The method of achieving additional spatial or temporal detail should be briefly explained. A map should be provided that shows the geography of any window or subarea relative to the whole region.

Traffic volumes, travel times, speeds and performance measures outside of the subarea are of very little interest and should not be reported unless there is a compelling need.

If a windowed network is exhibiting incorrect volumes at the edges of the network as a consequence of deleting highway segments just outside the window, then it is permissible to further focus upon those areas nearest the project when reporting results.

Guidelines for Time Series Models

Reports of statistical model estimations and forecasts should contain these elements, at a minimum:

  • List and description of data sources;
  • Any data cleaning to remove anomalies;
  • Any required interpolation because of missing data;
  • Brief mention of any associated techniques (e.g., smoothing and transformations) that would aid understanding of how the analysis progressed.
  • List of variables and coefficients;
  • Confidence level used to select independent variables into the regression equation or minimum t-scores;
  • Adjusted R-square;
  • Standard error of the estimate;
  • Forecast(s), baseline and/or with optional scenarios.

There is no need to report individual t-scores for independent variables, so long as each t-score meets minimum requirements.

A graph of the traffic count data series may or may not be helpful in communicating results, depending upon data quality. Anomalies or missing counts may convey a sense of weakness in the database, even when the model is statistically strong.

A graph of the trend line, with additional lines to indicate confidence intervals or scenarios, may be an effective communications tool.

Details of the forecast that are of limited value to the public or that are unnecessary for archival purposes may be omitted. Such details include the software used, specifics of how the software was set up, or variables that were tried and rejected.