Research Design

the idea that we should accept the simpliest explanations of the data; what are the minimal concepts that can explain the data?

how well have I asked the question

how well can I generalize these data

• conceptual basis/theoretical foundations this is far more inferential and base on expert opinion
• "What might be going on here to cause the effects I am observing or predicting?

• history (anything that occurs during the study)
• maturation (change over time)
• testing (familiarity with pre-post assessment
• instrumentation (changes in the instrument or procedures over time)
• statistical regression (extreme score regress to the mean)
• selection bias (systematic difference btwn groups differenc of people who self-select)
• attrition (ppl who drop out are different from those who stay)
• combination of selection and other threats (multiple threats, especially selection X history bias or maturation bias)
• diffusion or imitation of treatment ("leaking" tx into control group)
• special controls or reactions of controls (kind of like the placebo)

• sample characteristics (if characterists of sample don't match the sample you cannot generalize)
• stimulus characteristics and settings (can this be generalized beyond the setting of the lab, clinic, etc.)
• reactivity of experimental arrangements (responding in a way to make self look a specific way rather than authentically)
• multiple tx interference (the effects obtained in the experiment may be due in part to the context or series of conditioning in which it was presented)

• alpha error - probability of rejecting the null hypothesis when it is true
• you say there is a differnce but there isn't

• Beta error
• probablity of accepting the null hypothesis whe it is false
• you say there isn't a difference, but there is

magnitude of difference between the two groups

• hypothesis that when something happens to one, something happens to the other
• positive - both go the same way
• negative - go opposite ways

when one variable directly influences the nature, direction of another variable

variable or process that explains how a variable produces particular outcome

a repeated measures correlation - gives information about the amount of variabilty in your DV is accounted for by the IVs

• knowledge
• competence
• volition

influences to types of subjects who participate in experiments

• samples of convenience
• volunteer
• attrition

• 1. provides order where there is variability
• 2. explain change
• 3. inform choice of moderators
• 4. aids in connection to practice

a term used in qualitative research referring to hypotheses that emerge from intensive observations

• 1. true experiments
• 2. quasi-experimental
• 3. case-control designs
• 4. qualitative

subjects are randomly assigned to conditions- gold standard for determining effectiveness of an intervention

conditions of true experiment are approximated but cannot randomize, often due to impossiblility or ethical concerns

chose subjects (cases) sho vary in the characteristic of interest

• cross-sectional - makes a comparison between groups at a given point in time (cohort effect [differences really due to varying group histories])
• longitudinal - makes a comparison over an extend (shows change, attrition)

sometimes these designs are combined

• age
• sex
• SES
• IQ

• pretest-posttest
• posttest only design
• Solomon four-group design
• factorial designs
• crossover design
• multiple treatmetn counterbalance design

• needs two groups; one receives tx one doesn't
• allows you to match subjects and assign randomly
• w/in group variability is reduced and allows for more powerful statistcial tests
• weakness: does not control for the possibility of testingXtreatment or pretest sensitization

needs two groups; no pretest

lack of pretest make it less used b/c there could have been pre-existing differences

• evaluates the effect of pretesting on the effects obtained with a particular intervention
• needs 4 groups
• 2 in pre/post; 2 in post only
• controls for prestest impact; controls for re-testing

looks at two or more variables unlike other designs

allows you to look at interaction effects (e.g., between sex and treatment)

• partway through the experiment, subjects cross over and receive the other condition
• allows for random assignment

cross over design for more than two condintions which takes greater planning in order to ensure random sequencing