A Total Error Approach for Validating Event Data.
Understanding how useful any particular set of event data might be for conflict research requires appropriate methods for assessing validity when ground truth data about the population of interest do not exist. We argue that a total error framework can provide better leverage on these critical quest...
| Publicado en: | American Behavioral Scientist Vol. 66; no. 5; pp. 603 - 625 |
|---|---|
| Autores principales: | , , |
| Formato: | Artículo |
| Publicado: |
Sage Publications Inc.
May2022
|
| Materias: | |
| Acceso en línea: | Ver este registro en EBSCOhost |
| fields | @attributes: recordID: 1 pdfLink: plink: https://search.ebscohost.com/login.aspx?direct=true&db=ssf&AN=156391425&site=ehost-live header: @attributes: shortDbName: ssf uiTerm: 156391425 longDbName: Social Sciences Full Text (H.W. Wilson) uiTag: AN controlInfo: bkinfo: jinfo: jid: 00027642 ABS jtl: American Behavioral Scientist issn: 00027642 maglogo: Y pubinfo: dt: May2022 vid: 66 iid: 5 pid: 344 pub: Sage Publications Inc. artinfo: ui: 156391425 10.1177/00027642211021635 ppf: 603 ppct: 22 formats: tig: atl: A Total Error Approach for Validating Event Data. aug: au: Althaus, Scott Peyton, Buddy Shalmon, Dan affil: Cline Center for Advanced Social Research, 14589 University of Illinois at Urbana, Champaign, IL, USA su: Nigeria Petrarca, Francesco, 1304-1374 Boko Haram (Organization) Benchmark problems (Computer science) sug: subj: Nigeria Boko Haram (Organization) Benchmark problems (Computer science) Petrarca, Francesco, 1304-1374 keyword: event data gold standard ground truth total error paradigm validity event data gold standard ground truth total error paradigm validity ab: Understanding how useful any particular set of event data might be for conflict research requires appropriate methods for assessing validity when ground truth data about the population of interest do not exist. We argue that a total error framework can provide better leverage on these critical questions than previous methods have been able to deliver. We first define a total event data error approach for identifying 19 types of error that can affect the validity of event data. We then address the challenge of applying a total error framework when authoritative ground truth about the actual distribution of relevant events is lacking. We argue that carefully constructed gold standard datasets can effectively benchmark validity problems even in the absence of ground truth data about event populations. To illustrate the limitations of conventional strategies for validating event data, we present a case study of Boko Haram activity in Nigeria over a 3-month offensive in 2015 that compares events generated by six prominent event extraction pipelines—ACLED, SCAD, ICEWS, GDELT, PETRARCH, and the Cline Center's SPEED project. We conclude that conventional ways of assessing validity in event data using only published datasets offer little insight into potential sources of error or bias. Finally, we illustrate the benefits of validating event data using a total error approach by showing how the gold standard approach used to validate SPEED data offers a clear and robust method for detecting and evaluating the severity of temporal errors in event data. pubtype: Academic Journal doctype: Article src: R language: English refInfo: copyright: @attributes: flag: N holdings: @attributes: islocal: N |
|---|