Lecture 8 notes Practical research design Participants samples: Who will participate in my experiment? Who can I generalize my results to? External validity concern How large should my sample be? Random assignment Statistical power Smaller effects require much larger samples If r =. 8, then N=11 If r =. 2, then N=226 (study a relationship much stronger) Independent variable How strong should my manipulation be? Size o effects we observe is a ratio of manipulation strength to noise Stronger manipulations are easier to see (planet size) Controlled experiments reduce noise (lens clarity) Recommendation: Start w/strong manipulations More likely to determine if your hypothesized relationship is present Follow up with more subtle operational definition of IV What will my manipulation look like? Straightforward manipulations: Simple showing participants research stimuli Survey research, most computer-based research Advantages: simple, cost-effective,; deception not needed Staged manipulations: Elaborate "play" to generate psychological situations Advantages: experimental realism: Ss experiment the situation as engaging Generalizability beyond lab will usually be high, good external validity Straightforward manipulation example: rate each of these people according to how trustworthy they are: Stages manipulation example: Does competition enhance motivation? I was a confederate Memorized lengthy script
Staged reaction to losing Experimental realism seemed high Mundane realism, or similarity to real situations was not so high Dependent variables: Self-reports Behavioral Physiological Different measures impose different psychological constraints Recommendation use multiple methods/measures Is the relationship true under different conditions? Dependent variables: Different measures impose different psychological constrains Rate these faces quickly Memorize a number while rating these faces Rate these faces while inverted Pilot studies Primary use: debugging Are there logistical problems you did not foresee? Is the data you will be collecting useful? Is the psychological experience as you expect? Not: Is the hypothesis correct/working? (stop asking correct or wrong question) Demand characteristics: participants awareness of hypotheses generally leads to compliance (e.g .: goes exercising increase happiness? IV: do noy exercise vs. exercise 7x this week DV: self-reported happiness Is this a confound? Rules: Confound variable varies with level of the independent variable Confound variable is casually related to DV Solutions to demand characteristics: Between-subject designs:
Participants now only see one experimental condition Not foolproof, especially when working with highly intelligent and social participants Filler: additional experimental procedures items, or delays that obscure the hypothesis Placebo or expectancy manipulations Create identical psychological experience, with and without the active ingredient. Blinding: fully obscuring either hypotheses and/or the experimental manipulation Single-blind design: participants are unaware of their condition (e.g .: placebo vs. treatment) Double-blind design: both participants and experimenter are unaware