© 2001 dr. laura snodgrass, ph.d.1 conducting experiments choosing methods sampling and sample size...
TRANSCRIPT
© 2001 Dr. Laura Snodgrass, Ph.D. 1
Conducting Experiments
• Choosing methods
• Sampling and sample size
• Independent variables
• Dependent variables
• Controls
• Debugging
© 2001 Dr. Laura Snodgrass, Ph.D. 2
Choosing Methods
• Laboratory experiments can be artificial– too much control
• Field experiment– more natural setting– lose some control
• Ethical and practical concerns
• Participant variables– quasi-experimental designs
© 2001 Dr. Laura Snodgrass, Ph.D. 3
Choosing
• Description and prediction can be done without casual concerns
• Human complexity– number of interacting causal variables
• Neglect of individual differences– averaged across groups
• Social responsibility– objectivity– values– Gergen’s paradigm II
© 2001 Dr. Laura Snodgrass, Ph.D. 4
Sampling
• Generalization requires adequate sampling
• Populations– you define population “of interest”
• Why sample– cost-benefit analysis - law of diminishing returns– destruction of items tested– infinite populations– may increase accuracy
© 2001 Dr. Laura Snodgrass, Ph.D. 5
Participant Sampling
• Psychology as the study of white rats and college sophomores
• Generalization from– different species– different groups
• students have different pressures and performance anxiety
• volunteers differ from non-volunteers
© 2001 Dr. Laura Snodgrass, Ph.D. 6
Sampling Techniques
• Systematic random sampling
• Stratified random sampling
• Cluster sampling
• Haphazard or convenience sampling
• Quota sampling
© 2001 Dr. Laura Snodgrass, Ph.D. 7
Other Types of Samples
• Experimenters as samples– gender, age, ethnicity, behavior, dress
• Stimulus sampling– representative of pop of stimuli– random or controlled
• Condition sampling
• Response sampling– number of dependent measures– number of trials
© 2001 Dr. Laura Snodgrass, Ph.D. 8
Sample Size
• Tradition– look in journals
• Expected variability in results– consistency within and between participants
• Planned statistical analysis– parametric versus nonparametric– significance level– size of difference between means expected– do a power analysis
© 2001 Dr. Laura Snodgrass, Ph.D. 9
Independent Variables
• Setting the stage– informed consent– brief explanation of what is expected
• Types of manipulations– straightforward– staged
• to create a psychological state• to simulate a real world situation• use confederates
© 2001 Dr. Laura Snodgrass, Ph.D. 10
Independent Variables
• Strength of manipulation - choosing levels– number of levels– range– how close together are the levels
• Combining variables– incomplete or unbalanced designs (leaving out some cells)
• redundant or illogical• data from literature• too many cells to fill
© 2001 Dr. Laura Snodgrass, Ph.D. 11
Independent Variables
• Confounding– environmental confounds
• the “Hawthorne Effect”– participant confounds
• equality of groups
• Cost of manipulation
© 2001 Dr. Laura Snodgrass, Ph.D. 12
Dependent Measures
• Types of measures– self-report
• rating scales– behavioral
• reaction time• error rate
– physiological• GSR, heart rate
© 2001 Dr. Laura Snodgrass, Ph.D. 13
Dependent measures
• Sensitivity– ceiling effect - too easy– floor effect - too hard– no effect
• Multiple Measures– e.g. time perception
• Ethics of measures (e.g. privacy)
• Cost
© 2001 Dr. Laura Snodgrass, Ph.D. 14
Other Controls
• Participant Effects– loss of subjects– volunteers– social desirability– demand characteristics
• deception• filler items• placebo groups
© 2001 Dr. Laura Snodgrass, Ph.D. 15
Controls
• Experimenter effects– experimenter bias or expectancy effects
• subtle coaching• recording errors• teacher expectancy
• Solutions– training– run conditions simultaneously– single-blind– double-blind
© 2001 Dr. Laura Snodgrass, Ph.D. 16
Debugging
• Research proposals– getting feedback from others
• Pilot studies
• Manipulations checks– especially in pilot study– can explain non-significant results
© 2001 Dr. Laura Snodgrass, Ph.D. 17
Data Defects
• Missing data– some statistics allow for missing daat– can replace with averaging techniques– SPSS have several missing data options
• Extreme score or outliers– techniques for discarding– replace with averaging
• Appropriate Statistics!!!!!!!!!!!!!!!!!