Date

20 May 2016

Attendees

Agenda

1) Use Case reminder.

2) Where we are on our road map.

3) Open Action Items

4 ) JIRA Issues Review

https://jira.edmcouncil.org/browse/IND/?selectedTab=com.atlassian.jira.jira-projects-plugin:issues-panel

5) Todays content discussion.

UML

Spreadsheet

Protégé

6) For next week.

Proceedings

Slides reviewed on this week's telecon –

The primary discussion this week was on how to model patterns that the various indicators BLS reports fit into. Initially, we modeled employed person and unemployed person as children of person, which would be disjoint from one another and from "person not in the workforce" once that's added (see slides). But, based on today's discussion, that may or may not be right.

We agreed that we should attempt to model the primary concepts involved in these statistical measures – the notion of a population, such as an "employed population", an "unemployed population", and a "population not in the workforce", which are probably more appropriate than attempting to model these as kinds of individual people. The publications released by the BLS and other statistical agencies typically represent sample surveys of some subset of a population rather than an aggregation of reports received that reflect actual numbers for every person in their jurisdiction.

From a BLS perspective, the most abstract models talk about the 'thing being measured', as a unit type, and then refine that to reflect a particular area of interest, or universe, such as people, households, establishments, purchases, and so forth. For example, the universe specific to reports concerning employment and unemployment involves civilian, noninstitutional adults and minors that are 16 and older, in other words, the working population. The population for employment and unemployment numbers for a specific BLS release such as the employment situation summary for households is based on the universe for that report (the working population, broken down by employed, unemployed, and not in the labor force), time, and geography. The high level pattern reflects a primary actor, some utility (resources, things on hand, capabilities – the population sampled with respect to that utility), and time / event. For the producer price index, the population is limited to the producers / manufacturers, excluding other kinds of service providers.

Some additional specifics with respect to time:

the notion of a "month" in these reports does not reflect at what point during the month a person entered into a "state" – employed, unemployed or otherwise
it means that during the reporting month, a person moved from their prior "state" into that new "state" for at least one day
it is essentially a discrete measurement taken sometime during the month, similar to the U.S. census

An important thing to keep in mind is that these sample surveys are intended to be comparable over time

there is the notion of some discrete time period / instant used for comparison with prior periods
reports include current numbers and/or % change from one period to another
there is no notion of flow, however, from period to period, with respect to how the statistics should be interpreted

We also discussed the plausibility of estimates, and potential for comparison of the aggregate data used by the BLS with commercial resources. So for example, in 20 years time, the CPI may incorporate commercial data by region to get better estimates of things like the price of milk, purchasing patterns, etc.

Decisions

Elisa will continue attempting to model the higher level pattern using employment / unemployment as examples, and Dan will send a paper that may help in teasing this out over the coming week, with exchange primarily by email (Dan will be in Norway this coming week and Elisa will be at No Magic World, so meetings are unlikely). Elisa and Pete may use the time this coming Friday to work on this modeling pattern, rather than hosting the more general IND call (on Friday, 5/27).

Action items