Research Methods University of Massachusetts Boston ©2012 William Holmes 1 AB ?
Correlation is not Causation Post Hoc fallacy “For Example” is not proof Belief is not theory 2
Supporting Evidence Temporal Ordering Theory Rejection of Rival Hypotheses 3
Correlation between events Correlation between characteristics Data a result of research having internal validity (not biased research) 4
Potential outcomes Observed outcomes Individual versus average (expected) outcomes Completers versus non- completers Stable Unit and Treatment Value Assignment--SUTVA 5
Cause precedes effect Chronological ordering—happenstance Conceptual ordering—theory Process ordering—empirical process 6
Theoretical Assumptions of Program Social Psychological Economic Biological 7
Controlling Competing Influences— confounding variables Control by Sampling--exclusion Control by Design--randomization Control by Statistical Procedures-- controlling 8
V=Violence SA=Substance Abuse D=Depression 9 The rival hypotheses: violence causes depression vs. substance abuse causes depression
V=Violence SA=Substance Abuse D=Depression 10 Indirect effects between violence and depression
V=Violence SA=Substance Abuse D=Depression 11 Spurious correlation between violence and depression
V=Violence SA=Substance Abuse D=Depression G=Genetics Effect is conditional on other variable 12 G
History—external to experiment Maturation—developmental change Instrumentation—biased measures Subject-Experiment Interaction—persons messing with the experiment Testing Effects—persons learning from the experiment 13
14 Level of Measurement Measures of Association Test of Significance Nominal/categoricalPercentage, Odds Ratio, Logit, Phi t, Chi-square MixedPercentage, Means, Eta t, F Interval/RatioCorrelation, Regression t, F
Determine presence of relationship-- significance test or substantive criteria Determine strength of relationship-- measure of association Compare bivariate results with what happens when controls are introduced 15