Views of Transparency from AAPOR and OMB

Benefits and Costs of Transparency: Views from Agencies

Foreign View of Benefits and Costs, and Approaches Used for Transparency

Lunch

Eric Rancourt (Statistics Canada)

Case Studies: What Goes on Now Regarding Transparency?

What Does Transparency (and Reproducibility) Mean Operationally?

Internal to an Agency—Reproducibility Technology Transfer

Input Data Sources (Survey Data, Administrative Records Data, Other Sources)
Ability to Access Internally Archived Datasets Used to Generate Major Releases; Retention of Paradata to Enhance Value of Archived Survey Data; Assessment of Quality; Error Profiles; Retention of Paradata for Administrative Data—What Does That Mean? Quality Profiles for Administrative Data; the Use of Surveys for Assessing the Quality of Administrative Data; Standards for the Use of Administrative Records for Statistical Purposes; Some Facility for Experts to Operate Internally

Documentation of Methods
Can Get Much More Detailed Since Internal SAS; R Heavily Commented Code;
Can Get into Current Editing; Imputation, Weighting, etc.; Procedures Being Able to Recreate the History of Changes to These Algorithms over Time

External to an Agency—Building Trust, Benefits from Crowd Sourcing

Input Data Sources
PUMS, Anything else? PUMS for Administrative Records? Research Data Centers?

Documentation of Methods
Summary for General Analyst versus Availability of Commented Code But What to Run It On?
Not for Editing, Imputation but for Some Workflow History? How Crowd Sourcing Would Operate; How to Evaluate Suggestions; Very Time Intensive; Some Governance Necessary

Making Confidential Data Part of Reproducible Research

What Does Reproducibility Mean for Federal Statistics: An Academic’s Perspective

Adjourn

Improving Transparency of Statistical Data with Standards for Metadata and Work Processes

Building Trust: Summary of Workshop
Topics: Benefits Due to Crowd Sourcing; Need for Governance; What Tools Look Most Promising? Outreach to the Public; Outreach to Analysts; Transparency Will Bring Criticism (valid and invalid); Will Demand Additional Resources Both to Create the Archival and Documentation and to Interact with the Users; What Is Feasible and What May Not Be?

Panel Discussion:
Hermann Habermann (CNSTAT)
Tom Louis (Johns Hopkins University)
John Abowd (U.S. Census Bureau)
Bill Eddy (Carnegie Mellon University)

Working Lunch to Continue Morning Discussion

Adjourn