Using the specification curve to teach spatial data analysis and explore geographic uncertainties

Peter Kedron, Matthew Quick, Zach Hilgendorf, Mehak Sachdeva

Research output: Contribution to journalArticlepeer-review

Abstract

Educational materials focused on spatial data analysis often feature mathematical descriptions of methods and step-by-step instructions of software tools, but infrequently discuss the set of decisions involved in specifying a statistical model. Failing to consider model specification may lead to specification searching, or the process of repeating analyses to obtain results that meet the criteria thought to be required for publication, and the disproportionate reporting of false-positive results in the academic literature. This article proposes that the specification curve–a meta-analytical technique that visualizes the specifications and results from a large set of justifiable and plausible statistical models–be used as a pedagogical tool to teach (spatial) data analysis and explore the geographic uncertainties that arise when specifying and interpreting spatial regression models. An example specification curve that focuses on two common specification decisions in a spatial regression model, specifically selecting predictor variables and constructing the spatial weight matrix, is illustrated. Strategies for using the specification curve in educational contexts to develop analytical plans, reflect on the generalizability of research findings, and highlight issues of replicability and publication bias are proposed.

Original languageEnglish (US)
JournalJournal of Geography in Higher Education
DOIs
StateAccepted/In press - 2021
Externally publishedYes

Keywords

  • Model specification
  • geographic information science
  • geography education
  • regression
  • spatial analysis
  • uncertainty

ASJC Scopus subject areas

  • Geography, Planning and Development
  • Education

Fingerprint Dive into the research topics of 'Using the specification curve to teach spatial data analysis and explore geographic uncertainties'. Together they form a unique fingerprint.

Cite this