Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter.
Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.
OCR for page 166
P R O P O S E D VA L I D AT I O N A N D S E N S I T I V I T Y T E S T I N G O F D E N V E R R E G I O N 155
acceptable error ranges will be determined by estimating those produced using the four-step model. While this may
the variation expected in aggregate model statistics (such seem to be a rather lenient standard, it must be remem-
as vehicle miles traveled [VMT]) resulting from the devi- bered that the four-step model was, in fact, calibrated to
ation on the particular model component, in effect tying produce reasonable validation results for 1997 and 2001.
new validation tests that as yet have no standards to One would expect that, because the ABM can consider
existing tests that do have such standards. more aspects of personal activity performance and travel
Sensitivity testing will also play an important role in behavior, the amount of "adjustment factoring" not tied
the validation of the ABM components. There are no to specific measurable behavior should be less in the ABM.
established standards for reasonable elasticities for some Another validation activity under consideration is
of the newer model components, so this will be more of "back-casting" to a prior year (besides the model esti-
a reasonableness test than anything else. In such cases, it mation year of 1997). This will be done if the necessary
may not be possible to tie these tests to existing tests with data are available and the resources to perform the back-
standards already developed. In such cases, it may be cast are available.
possible to compare Denver model test outcomes with In addition to specifying traditional model validation
observed outcomes in other cities with conditions similar standards, input and coordination with federal agencies
to those being evaluated in the sensitivity tests. will be sought in the validation and sensitivity testing of
Significant effort will be placed on the validation of the ABM for the Denver region. This will be particularly
the overall model system. Again, this will be comparable important in the development of the ABM because oper-
to previously recommended validation procedures and ational experience with them is limited; federal agencies
will include the following: may be expected to evaluate them closely for validity
and for consistency with the calibration outcomes of the
· Reasonableness and logic checks of demographic numerous trip-based models in existence. DRCOG
and network data or skim data input to the models; intends to involve federal agencies early in the process,
· Traditional validations for the model estimation through oversight panels or other means, and will
year (1997) and for 2005 against independent observed include its requirements in the calibration and validation
data. Depending on data availability, these traditional plan at the earliest point.
checks will include:
Root-mean-square error of modeled to observed
traffic volumes by appropriate segmentation vari- TEMPORAL AND POLICY SENSITIVITY TESTING
ables (such as facility type, traffic volume level, and
so forth) The normal validation testing outlined above includes
Matching regional observed VMT within some temporal validation in that the model will be vali-
approximately 1% error dated against observed travel data for 1997 and 2005.
Matching observed VMT by facility type Such testing is crucial for model validation but does not
Matching VMT by area type address the hypothesized true value of ABMs--the pro-
Matching total transit boardings duction of better impact assessment information and
Matching transit boardings by sub-mode travel forecasts that will result from the more appropri-
Rapid transit boardings by corridor, sub-mode, ate representation of the actual decision process.
and station Two approaches will be used to test the sensitivity of
Park and ride lot usage the ABM. The first will be the application of the ABM
Matching a series of at least 10 highway and for an existing forecast year and scenario and compari-
transit screenline volumes son of those results to those produced by the calibrated
Highway volumes on individual freeways four-step model. While the "true" results for such a fore-
Toll road usage cast year cannot be known, the results from the four-step
Acceptable matching of peak and off-peak model provide one outcome that has been deemed rea-
speeds sonable. Several questions will be asked:
Roadway speeds by several time-of-day periods
(a.m. versus p.m. versus off-peak, and so forth); and · How similar are the results? The traditional valida-
· Tests of the sensitivity of the overall model system tion measures outlined in the previous section regarding
to input variables (similar to the procedures used for the traffic volumes and transit boardings can be used to mea-
model component sensitivity testing). sure the similarity.
· Which model produces more believable results?
The ABM will be subjected to the same validation stan- Two outcomes are possible: the forecasts from the two
dards that were used for the recently refreshed four-step models are not substantially different or the forecasts are
model. Results are expected to be as good as or better than substantially different. In either case, an assessment will
