EXPERIMENTAL DESIGNS Criteria for Experiments Independent, Dependent, and Confounding Variables Types of Experimental Designs Threats to Internal Validity Threats to External Validity
Criteria For Experimental Designs Cause: experimenter manipulates a variable Comparison: more than one condition Control: extraneous variables are eliminated
Independent Variable variable manipulated by the experimenter levels conditions
Dependent Variable variable measured to assess the effect of the independent variable
Confounding Variable variable other than IV and DV which changes between conditions control variable: potential confounding variable that is controlled
Types of Experiments between subjects matched groups within subjects
Threats to Internal Validity: Individual Differences Systematic differences between individuals in different groups Strategies random assignment matched groups within subjects design
History Events outside the experiment Most likely when conditions are measured at different times with long delays Strategies Decrease time between conditions Add a control group measured at same times
Maturation Physical changes related to aging Particular problem for within-subjects designs Strategies decrease time between measurements add a control group measured at same times
Instrumentation Changes in the measuring instrument or equipment Strategy Use standardized administration
Attrition Participants drop out of the study at different rates for the different conditions Strategies Check attrition rates across groups Compare participants who drop out to those who stay in
Diffusion of Treatment Information about the purpose of the study is shared with future participants Strategies Short time span between participants Use debriefing to request that participants do not share information about the study
Demand Characteristics Cues from the experimenter or research procedure about what behavior is desired Strategy Single-blind procedure
Experimenter Effects Experimenter’s expectations affect measurements Strategy Double-blind procedure
Floor and Ceiling Effects Measuring instrument is not sensitive enough Floor effects Ceiling effects Strategy Check sensitivity of instrument prior to experiment
Regression to the Mean When measured twice, scores on the second testing tend to be closer to the mean Statistical phenomenon due to chance Strategy Don’t select participants for groups based on extreme scores Use an equivalently selected control group that does not get the treatment
Order Effects Also called Testing or Repeated Testing Effects of repeated measurements Fatigue effects Practice effects Carryover effects Strategy Counterbalance order of conditions
How Counterbalancing Works Change the order of conditions Order effects will still exist but will affect all conditions equally This prevents order effects from being confounding
Complete Counterbalancing Each possible order of conditions is used for an equal number of subjects If your conditions are A,B, and C, 1/6 of participants will get each order: ABC CAB ACB CBA BAC BCA
Latin Square Counterbalancing Each condition is presented in each position for an equal number of subjects Controls for practice and fatigue effects
Example Latin Square 1st 2nd 3rd 4th 1/4 get A B C D 1/4 get B C D A 1/4 get C D A B 1/4 get D A B C
Balanced Latin Square Latin square with additional requirement that each condition precedes and follows every other condition equally often Controls practice and fatigue effects Controls simple carryover effects (involving effect of a single condition)
Balanced Latin Square 1st 2nd 3rd 4th 1/4 get A B D C 1/4 get B C A D 1/4 get C D B A 1/4 get D A C B
Randomized Counterbalancing Used when there are multiple stimuli tested for each condition Put the stimuli in random order for each participant
Threats to External Validity Unrepresentative Sample use random or stratified random sampling do exact or systematic replications Artificiality use a more realistic setting do systematic or conceptual replications