How does one design an experiment and analyze the resulting data? Illustrative example: “Does viewing violence increase aggression?” Designing the experiment: Internal and external validity Controlling for error Summarizing the obtained data: Central tendency and variability Calculating a “test statistic” (Z) Interpreting the value Z
Research design terminology (from last week): Independent variable Dependent variable Systematic (confounding) error Unsystematic (random) error “Holding constant” Randomization Controlling for subject expectations: e.g. placebo control Controlling for experimenter expectations: e.g. double-blind control
Summary of hypothetical results from a correlational study relating amount of violent TV viewing and degree of “manifest aggressiveness” (shown earlier): Degree of “manifest aggressiveness” r > 0 Amount of violent TV viewing
Independent Variable: Research hypothesis: “Watching violent TV provokes aggression.” Experimental group Control group Independent Variable: Exposure to the experimental condition Exposure to the control condition Dependent variable: Performance Performance Is the difference statistically significant?
Graphical summary of data Violent Nonviolent Type of TV segment Mean seconds on “hurt” button 12 10 8 6 4 2 Girls Boys
(The 68 Girls) Experimental Control X (X-M)2 1. 11.0 5.76 1. 6.0 0.16 1. 11.0 5.76 1. 6.0 0.16 2. 8.2 0.16 2. 4.6 3.24 3. 12.3 . . 34. 8.7 13.69 . 1.40 3. 3.8 6.76 . . . 34. 6.4 0 X X
Numerical and Graphical Summary of Data from the 68 Girls: Mean: Variance: S.D.: Experimental group (N=34) Control group (N=34) 10 8 6 4 2 Mean seconds on “hurt” button Violent Nonviolent
Calculating Z from the data: (Z is the test statistic.) ?
Normal distribution of Z: .05 +1.65 (critical value of Z) Z