Originally presented to the 11th U.S. Networked Knowledge Organization Systems (NKOS) Workshop at the 2017 International Conference on Dublin Core and Metadata Applications, Washington, D.C.Recommended citation: Currier, B. D. & Butler, C. R. (2017). Research Data Reproducibility and the Importance of Attachment Level Metadata. Presentation to the 11th U.S. Networked Knowledge Organization Systems (NKOS) Workshop at the 2017 International Conference on Dublin Core and Metadata Applications, Washington, D.C. Accessed through LIS Scholarship Archive. Available at http://doi.org/10.17605/OSF.IO/7KUGAThough there are inconsistencies in the way that data reproducibility is currently defined within the social sciences, it is often used to mean simply that data and code are made available as a supplement to a primary object, such as a paper, and that these materials may be used to recreate identical results. However, Clemens asserts that a broader, more clearly defined range of ways in which data may be verified and reused, such as reanalysis and extension, is important in facilitating collaborative discussions that ultimately lead to better research. This presents a new curation challenge and a shift in the purpose of research data metadata as data and code themselves become primary research objects.Metadata elements either applied within a content management system or embedded within the object itself at the item, collection, or other hierarchical level in a digital collection (herein called attachment level metadata) is an important and often overlooked consideration for the purposes of research data management and reproducibility. Applying metadata at the highest possible level of attachment in a hierarchical object structure can optimize the schema and reduce redundancy (Sundgren, et al.). However, no matter how well-developed a metadata schema is, if an object becomes separated from the schema then it risks losing much of the contextual information necessary for broadly defined reproducibility. For this reason, a selective combination of embedded metadata and associated metadata at multiple hierarchical levels has the potential to be most effective.With this in mind, the Federal Reserve Bank of Kansas City is currently developing recommendations for file structure and organization, file formats, naming conventions, and metadata schema requirements for research data collections in preparation for implementing a research data preservation platform. These recommendations are based on international standards, such as the Dublin Core Metadata Initiative (DCMI) Metadata Terms, and industry practice, as ascertained from an internally-developed sampling of almost 250 economic journal policies created by cross-referencing journal impact factors, h5-indices, IDEAS rankings, and Federal Reserve Bank of Kansas City staff authorship and service to the journal (Butler and Currier). The various components of the recommendations intersect to support the overall usability, discoverability, interoperability, reproducibility, and preservation of research data as a primary object.This presentation will discuss the differences in and importance of both associated and embedded metadata at multiple levels of hierarchical attachment and the ways in which internal recommendations in these areas are being developed to optimize the reproducibility of research data.ReferencesButler, Courtney R., and Brett D. Currier. 2017. “You Can’t Replicate What You Can’t Find: Data Preservation Policies in Economic Journals,” Presentation at the 43rd IASSIST Annual Conference, Lawrence, KS, May 23-26.Clemens, Michael A. 2017. “The Meaning of Failed Replications: A Review and Proposal,” Journal of Economic Surveys, vol. 31, no. 1, pp. 326 – 342. Available at https://doi.org/ 10.1111/joes.12139Sundgren, B., Thygesen, L., and Denis Ward. 2008. “A model for structuring of statistical data and metadata to be shared between diverse national and international statistical systems,” OECD Working Paper. Available at http://www.oecd.org/std/38541998.doc