copy: Chapter 10: Using Between-Subjects and Within-Subjects Experimental Designs (copy)

Experiment:

  • Randomly sampling: choosing participants from a population - external validity

  • Randomly assigning subjects: randomly assigning participants from the population into groups - internal validity

Quasi-experiment:

  • Can't or don't randomly assign subjects

Experimental Designs

  • Between-subjects: each treatment administered to a different group (ex., comparing a control group and a treatment group)

    • Different groups of subjects in different groups

      • Ex. hot room participants vs cold room participants

  • Within-subjects: one group is exposed to all treatments (ex., one group takes a pretest and a posttest, and their pretest scores are compared to their posttest scores)

    • One group with 2+ variables

      • Ex. Tiffany's score in the hot room vs her score in the cold room

    • Cons

      • Practice effect - create strategies

  • Single-subject: one participant exposed to all treatments

Problem of Error Variance

  • Remember error variance (discussed earlier)

    • Ex. SAT scores = prep or not did affect scores

      • Ability to deal with stress, previously taken it before, quality of school, IQ, room temperature, etc. = error variance

    • Extraneous variables: variables not considered in the study; sometimes, you can control them. 

      • The more things you can make the same, the better. 

    • Subject variables: the difference between people; if you feel tired, your mood, if you do word searches a lot, 

      • To prevent, give instructions, match people, compare yourself to yourself (within subject design)

    • Variance in the

Handling Error Variance

  • Reducing Error Variance

    • Within-subject design (not always)

  • Increasing the Effectiveness of Your Independent Variable (I don't understand)

    • ABA therapy vs sensory integration therapy

  • Randomizing Error Variance Across Groups

    • Control for h

  • Within-subjects designs decrease error variance by controlling for subject characteristics

  • still have extraneous variables as a source of error variance

Within-Subjects Designs

  • Single-Factor Two-Level Design [2 options for how to do this] 

  • Option 1:

    • Randomly select participants from the population of interest

    • Expose all participants to 1 level of the IV

    • Measure the DV

    • Expose all participants to the other level of the IV

    • Measure the DV

    • Run a dependent samples t-test to determine if there are differences in the DV across levels of the IV

    • Ex: caffeine (2 levels) and concentration (continuous)

  • Option 2- Counterbalancing:

    • Randomly select participants from the population of interest

    • Randomly assign participants to 2 groups

    • Expose participants in group 1 to one level of the IV and participants in group 2 to the other level of the IV

    • Measure DV

    • Swap group exposure (expose participants in group 1 to the second level of the IV and participants in group 2 to the first level of the IV

    • Measure the DV

    • Run a dependent samples t-test to determine if there are differences in the DV across levels of the IV

    • Ex: caffeine (2 levels) and concentration (continuous)

Correlation design

  • Randomly sample from the population of interest

  • Have them fill out measures on at least 2 different variables

  • Run the correlational analysis

    • Ex. hypothesis correlation between room temperature and grades on a concentration test

      • The hotter it is, the higher the grade - positive correlation

key terms:

  • between-subjects design

    • An experimental design in which different groups of subjects are exposed to the various levels of the independent variable

      • Ex. study about sleep & memory. Group 1: 8hrs of sleep; Group 2: 4hrs of sleep; memory recall is compared

  • within-subjects design

    • An experimental design in which each subject is exposed to all levels of an independent variable

      • Ex: all participants get tested on their reaction time after drinking coffee and after drinking decaf

  • single-subject design

    • An experimental design that focuses on the behavior of an individual subject rather than groups of subjects.

      • Ex: therapists track one child over weeks to see if the new behavioral intervention is effective

  • Error variance

    • Variability in the value of the dependent variable that is related to extraneous variables and not to the variability in the independent variable.

      • Ex: in a study about stress, the participants' initial anxiety level adds error variance to the results

  • randomized two-group design

    • A between-subjects design in which subjects are assigned to two groups randomly.

      • Ex: participants get randomly assigned to 2 groups. 1 is exposed to CBT, and the other is not, then depression scores are compared.

  • parametric design

    • An experimental design in which the amount of the independent variable is systematically varied across several levels.

      • Ex: study on the impact of caffeine; different groups receive different amounts of caffeine.

  • nonparametric design

    • An experimental research design in which levels of the independent variable are represented by different categories rather than different amounts.

      • Ex: study on problem-solving skills across different types of learning (visual, auditory, or kinesthetic)

  • multiple control group design

    • Single-factor, experimental design that includes two or more control groups.

      • Ex: testing a new antidepressant; 1 group gets the drug, 1 gets a placebo, and one gets regular therapy

  • matched-groups design

    • Between-subjects experimental design in which matched sets of subjects are distributed, at random, one per group across groups of the experiment.

      • Ex: before testing a new teaching method, students are matched on GPA, then randomly assigned to the traditional method or the new method.

  • matched-pairs design

    • A two-group matched groups design.

      • Ex: 2 participants with similar IQ are paired, then one does mindfulness training and the other doesn't

  • carryover effect

    • A problem associated with within-subjects designs in which exposure to one level of the independent variable alters the behavior observed under subsequent levels.

      • Ex: in a mood study, participants who listened to happy music first might stay cheerful, which can affect their performance after listening to sad music

  • Counterbalancing

    • A technique used to combat carryover effects in within-subjects designs. Counterbalancing involves assigning the various treatments of an experiment in a different order for different subjects.

      • Ex: half of the participants listen to happy music while the other half listens to sad music to balance order effects

  • factorial design

    • An experimental design in which every level of one independent variable is combined with every level of every other independent variable.

      • Ex: study on caffeine (none vs 200mg) and sleep (8hrs vs 4hrs). 2x2 factorial

  • main effect

    • The independent effect of one independent variable in a factorial design on the dependent variable. There are as many main effects as there are independent variables

      • Ex: from the example above, if the group with 8hrs of sleep performed better regardless of caffeine, sleep is the main effect

  • Interaction

    • When the effect of one independent variable on the dependent variable in a factorial design changes over the levels of another independent variable.

      • Ex: from the example above, if caffeine only improved alertness with 4hrs of sleep, there is an interaction between sleep & caffeine

  • simple main effect

    • In a factorial analysis of variance (ANOVA), the effect of one factor at a given level (or combination of levels) of another factor (or factors).

      • Ex: from the example above, analyzing the effect of caffeine only for participants who slept 4 hrs would test a simple main effect

  • higher-order factorial design

    • An experimental design that includes more than two independent variables (factors).

      • Ex: A psychologist tests how sleep (4 vs. 8 hrs), caffeine (none vs. 200 mg), and music (quiet vs. upbeat) together influence concentration.  2×2×2 design.