Disaggregation II

 

Breaking Up Is Hard to Do

They say that breaking up is hard to do
Now I know
I know that it's true
Don't say that this is the end
Instead of breaking up I wish that we were making up again

Disaggregating is breaking up.

I am returning to the original purpose of this blog, which was to keep from repeating my self to colleagues, and instead referring them to a link to a blog post. So while the subject probably does not interest the vast majority of you, here it goes.

1.  Define disaggregation. The unique primary key fields in the Freight Analysis Framework, FAF, are Origin, Destination, Commodity, and Mode. Disaggregation is the opposite of aggregation. Each of these reported primary key fields are an aggregation.

• Origins and destinations are reported at level of FAF regions. Domestically these are states or portions of states, and they can be disaggregated to county, districts, firms, etc. However, I do not believe there is an interest in disaggregating International FAF regions into countries. (e. g. Europe into Germany, France, etc.).
• Commodities are in the Standard Classification of Transported Goods at the two-digit level, SCTG2 system and while these could be either disaggregated into another commodity classification system, e.g. Standard Transportation Commodity Code, STCC, or be disaggregated within the hierarchical SCTG system, e. g. SCTG5,  I do not believe that this is what is meant by disaggregation.
• Similarly, modes have been aggregated, and can be disaggregated ( e.g. Water into Great Lakers, Inland barges, Deep Sea ships), but I do not believe that this is the intent.

IMHO disaggregation is intended to mean only DOMESTIC origins and destinations into smaller zones. 

2. Why are you disaggregating. IMHO, disaggregation is done of the trip table for use in analysis. And disaggregation is also done of the trip table for use in traffic assignment. When it is done for the trip table to be used in analysis, such as the preparation of Freight Plans, assignment may, or may not, also be done. When it is done only for assignment, to prevent “lumpy loading,” i.e. over assigning links, and the link volumes, and not the trip table itself, may be analyzed. So disaggregation does NOT mean assignment. In fact, the assignment generally is done of vehicles that carry freight while the FAF is the contents of those vehicles. These are NOT the same. 

3. What does disaggregation include? To STEM-ers like myself, disaggregation is simply the opposite of aggregation, which is if you compress a matrix to aggregate it, then you expand a matrix to disaggregate it. This means that if the matrix has II/IE/EI/EE, and I is the internal geography to be disaggregated and E is the external geography not to be disaggregated, then EE is NOT disaggregated/expanded. However non-STEM-ers often include EE pass through as part of disaggregation. It is possible to identify EE pass through, it is simply subarea extraction, but it needs a trip table AND a network, unlike matrix expansion which only needs a trip table. So, does disaggregation include, or exclude, EE pass through? 

4.  Where do you get the disaggregation factors? If you have decided to disaggregate Domestic Origins and Destinations, it is important to understand the reported Domestic Origins and Destinations in the FAF are sometimes POEs, i.e. Ports of Entry for imports and Ports of Exit for exports, not origins and destinations.  In the FAF, actual origins and destinations are correlated with the economy of the shippers and receivers, but the POE is correlated with the carrier(s). There are different sources for economic disaggregation factors, and carrier disaggregation factors. You can infer what kind of a POE it is based on the reported foreign and domestic modes. 

5. How finely can you disaggregate? The FAF is an expansion of a sample, e.g. the Commodity Flow Survey, CFS, and additions that are Out Of Scope, OOS, with respect to the CFS. The expansion of a survey is for statistically significant zones. To protect privacy of the survey respondents, the original survey and the expansion factors are not reported. Beside even if the surveys were reported, they would NOT be statistically significant for smaller geographies. Additionally, for OOS there might not even be surveys and the whole concept of statistically significant does not apply. But you can come up with most probable disaggregation factors even if you can not come up with exact disaggregation factors. When disaggregating on the most probable you may be effectively betting on the favorite but that does not preclude the possibility that the factors associated with the long shot should have been used. Because these are most probable, not exact, Disaggregation Factors, I am uncomfortable disaggregating, inferring, too finely. 

6. How different should disaggregation factors be? The FAF is an asymmetrical table, e.g. the tons of grain from Iowa to New York City does NOT equal the tons of grain from New York City to Iowa. The FAF is asymmetrical across every primary key field, i.e. origin/destination/commodity/mode. Thus, ideally disaggregation factors should also be different by origin/destination/commodity/mode. But this is the ideal. The more dimensions that can be included the better, but some dimensions, and not others, is better than nothing. 

7. How should a successful disaggregation be judged? Is the goal to match the original table or to match some secondary source? What are those secondary sources and what is the error between them and the original FAF? The difference between the disaggregated FAF and the secondary source might in fact only be error between the original FAF and the secondary source. Matching the original if the disaggregation factors sum to 100% is easy, but this may not match the secondary source. 

In conclusion

·        What is meant by disaggregation? Does it include only II/IE/EI or does it also includes EE passthrough?

·        What is the deliverable? Is it the disaggregation process and a set of default disaggregation factors, or does it also include the process by which to create new disaggregation factors.

·        If the process of creating new disaggregation factors relies on public data sources, what is to be done if the format, contents, link location  or processing of the public sources which are not under the consultants control change  (e.g. US Army Corps of Engineers, USACE, Waterborne Commerce Statistics Center, WCSC, changed the format, contents, and location of the reporting that is usable with FAF3 (2007), FAF4 (2012) and FAF5 (2017).  The number of FAF regions is different in FAF3 from those in  FAF4 or FAF5. The number of regions stayed the same between FAF4 and FAF5, but three counties switched regions. Also, the numeric code associated with Rhode Island as a FAF region changed from 440 to 441. (not disputing either change in the FAF, only pointing out that it did change, and this has impacted the software). The counties not served by active rail tracks or navigable water appears to be no longer reported by Oak Ridge National Laboratory, ORNL, Center for Transportation Analysis, CTA.