COMM 88 Lecture Notes - Lecture 5: Internal Validity
4/17/18
●Casual relationships between variables
○X influences/affects/changes y
●Different methods for testing different relationships
○survey/observational research (researcher A)
■Tests associations (just relationships/correlations)
■measure/observe some attitudes/behaviors & correlate them or compare existing
groups of people on some measure
●Great for external validity - ability to generalize results to other people
(if use representative sample) and to normal life settings (if observe or
ask people about normal behavior, etc)
●Poor for causality
○Experimental research (researcher B)
■Tests casual connections: manipulate variables, separate people into groups and
give different “treatment” to each group; control everything else and measure
effects
●Great for internal validity - ability to estimate that x causes y
●Not just connection between variables but also estimate time order
(which variable came first)
●Rules our extraneous (3rd) variables/causes
●Poor for generalizability
●Defining concepts and Variables
○Independent Variable (IV)
■In experiments: a “causal” variable (the cause in cause/effect relationship),
manipulated by researcher
■In surveys/observational studies: a “predictor” variable (predict does NOT mean
“cause”)
○Dependent variable (DV)
■In experiments: an “effect” or outcome, the variable affected/changed by the IV
■In survey/observational studies: a variable being predicted by the IV
●Conceptualizing your variables
○Defining what the concepts mean for purposes of investigation
Document Summary
Measure/observe some attitudes/behaviors & correlate them or compare existing groups of people on some measure. Great for external validity - ability to generalize results to other people (if use representative sample) and to normal life settings (if observe or ask people about normal behavior, etc) Tests casual connections: manipulate variables, separate people into groups and give different treatment to each group; control everything else and measure effects. Great for internal validity - ability to estimate that x causes y. Not just connection between variables but also estimate time order (which variable came first) In experiments: a causal variable (the cause in cause/effect relationship), manipulated by researcher. In surveys/observational studies: a predictor variable (predict does not mean. In experiments: an effect or outcome, the variable affected/changed by the iv. In survey/observational studies: a variable being predicted by the iv. Defining what the concepts mean for purposes of investigation. Deciding exactly how the concepts will be measured (or manipulated) in a study.