Experimental and Nonexperimental Research Design
Chapter 7: An Introduction to Scientific Research Methods in Geography
- Daniel R. Montello and Paul C. Sutton
- Geography 4020- February 2nd 2010
ContentsEmpirical ControlCorrelation and CausalityLaboratory and Field SettingsBasic / Specific Research DesignDevelopmental DesignSingle-Case and Multiple-CaseComputational Modeling
Empirical Control in ResearchEmpirical Control – Any method of increasing
the ability to infer causality from empirical data
3 ways of exercising empirical controlPhysical ControlAssignment ControlStatistical Control
Experiment - Manipulation of VariablesNonExperimental - May involved physical or
statistical control…but no variable manipulation
Correlation is not causalityOr…. “Correlation is causality, but the
specific pattern of that causality is ambiguous.”
A B
A B
A BA B
C
A B
C D E F
(A) (B)
(C) (D)
Laboratory vs. Field Settings
(http://www.nd.edu/~druccio/images/frankenstein_lab.jpg)
(http://www.spatiallyadjusted.com/2008/08/05/breaking-the-tribe-mentality/)
Lab allows physical control while conducting studies.
Field settings allow researcher to examine a phenomenon where it normally occurs.
Basic Research DesignVariables are requiredGenerally 2 or more variables so a
relationship can be examinedLevels of Variables
Between Case Sometimes unavoidable
Within Case Are more efficient Lead to higher precision Reduce confounds
Specific Research DesignsAssorted research designs (Table 7.1 pg.120)Posttest-only design vs. Pretest-posttest
designFactorial Design
Multivariable manipulationAllows investigation of factor interactions
A1 B1
A1 B2
A2 B1
A2 B2
2 variables with 2 possible options per variable
Developmental Designs (Δ/Time)Developmental Designs – studies designed to
conduct research on developmental processes.
2 basic approachesCross-sectional – comparing 2 or more
groups(cohorts) at different stages of development.
Longitudinal – a ground of cases at one level compared to itself over time
Sequential Design – a hybrid approachTemporal scale is important to consider at
design phase.
Single-Case and Multiple-Case DesignsSingle-case experiment – a repeated
measures design within a single case.Improve by returning to original condition
(reversal design)Nonexperimental Example: Case study
Multiple-Case DesignBetter idea of how results generalizeSignal vs Noise
Nomothetic and Idiographic approaches to knowledge
Computational ModelingComputational models are typically
instantiated as sets of equations and other logical/mathematical operations expressed in a computer programSimplified representation of realityModel output can be considered “Simulated
data” and are typically compared to standard empirical measurements.
Gives empirical access to events that would be otherwise very difficult or impossible to study.
Example of complex climate modeling.
Steps of Computational ModelingCreate conceptual modelCreate computational model
ID parametersRun the computer programCompare model output to empirically
obtained dataRefine model and repeat initial steps if
necessary with new insight.Accept, Use, and communicate model
(Summary of Table 7.2 p.130)
ReviewQuestions???
What are the 3 forms of empirical control in research?
What are Confounds?Describe the difference between within-case
and between case design. Pro/Con of approaches?
Discuss how computational modeling might help you better understand or design real world empirical research.