Research Methods: Experimental Design I (Single Factor)
Post on 25-May-2015
4247 Views
Preview:
DESCRIPTION
Transcript
Single-Factor Experimental Designs
Brian J. Piper, Ph.D.
Goals
• Single Factor: 2 levels– Independent groups– Matched groups– Non-equivalent groups– Repeated measures (2)
• Single Factor: 3+ levels• Control Group Designs
Really?
• “Single factor studies using only two levels are not as common as you might think. Most researchers prefer to use more complex designs, which produce more elaborate and intriguing outcomes.”
2 Independent Groups
• Independent variable (IV) manipulated• Equivalent groups created by randomization• Does insecurity increase materialism?– Participants: undergraduate students (N=60)– IV: • Control: write essay about listening to music• Experimental: write essay about own death
Kasser & Sheldon (2000) Psychological Science, 11, 348-351.
t-test
• Student’s (Between Subjects) t-testt = (Mean1 – Mean2)/Variability12 if SD1 ≈SD2
Degrees of Freedom = N - 2
O
William S. Gosset
1876-1937
Assumptions of t-test1) Data is interval or ratio2) Data is normally distributed3) For two-groups, homogeneity of variance
Logic
If p > .05 assumption is metIf p < .05 than assumption not met.
Overview
t = Signal / Noise
+ ∞ to - ∞
2 Independent Groups
• Dependent Variable (DV): economic value
Mean (SD)t (degrees of freedom) = # , p <> .05
Matched Groups Example
• Some variable is measured & then group assignment
• Sleep deprivation & suggestibility example– Sleep length recorded before assigning to
deprivation (+ or - )– Long-story presented following by leading questions
(Were assailants armed with gun or knife?)
Blagrove (1996) J Exp Psychol: Applied, 2, 48-59.
Matched Groups Example
• Some variable is measured & then group assignment
• Sleep deprivation & suggestibility example• Results: 43, but not 21, hour deprivation
increased suggestibility
Blagrove (1996) J Exp Psychol: Applied, 2, 48-59.
Non-Equivalent Groups
• He quit because his “body forgot the urge to smoke”
• Comparison of brain damage to insula (N=19) versus damage to other areas (N=50)
Comparing Groups
• Smoking following brain damage was examined in patients with or without insula damage.
Naqvi et al. (2007). Science, 315, 531-534.
Results
Naqvi et al. (2007). Science, 315, 531-534.
Repeated Measures
• If same subjects get both treatments, there is a concern about order effects (fatigue, practice, etc.)
Repeated Measures
• If same subjects get both treatments, there is a concern about order effects (fatigue, practice, etc.)
• Solution: Counterbalance!
Mental Effects of Steroids
• Men (N=56) received:– steroids, placebo– Placebo, steroids
Pope et al. (2000) Archives General Psychiatry, 57, 133-140.
Repeated Measures: Example 2
• Do toddlers use visual feedback for balance or only mechanical feedback?
Lee & Aronson (1974) Perception & Psychophysics, 15, 529-532.
Repeated Measures Design
• Within-Subjects: order alternating– Forward – Backward – Forward – Backward– Backward – Forward – Backward - Forward
Lee & Aronson (1974) Perception & Psychophysics, 15, 529-532.
Repeated Measures Design
• Within-Subjects: order alternating– Forward – Backward – Forward – Backward– Backward – Forward – Backward - Forward
Lee & Aronson (1974) Perception & Psychophysics, 15, 529-532.
Paired t-test• Comparison of Novel Faces & Places Response
Times (msec) for correct versus incorrect decisions
• Degrees of Freedom = N - 2
2-Level Design Statistic
Independent groups Two-sample t-testMatched groups Paired t-testNon-equivalent groups Two-sample t-testRepeated measures (2) Paired t-test
3+ Levels
• Provide more information than 2 levels
3+ Levels
• Provide more information than 2 levels
1850-1909
Hermann Ebbinghaus
Data Presentation
• Whenever possible figures • If you must …………..tables
3+ Groups Analysis
• Analysis of Variance (ANOVA)– Signal / noise– Range: 0 to infinity, ≈3.0+ is statistically significant
1
ANOVA
• Step 1: Is there a difference (somewhere)?
• Degrees of freedom:– between groups = # groups -1– Error = total N – dfBG -1
• Step 2: Where is the difference?– T-test– Post-hoc tests (Tukey HSD)
1915-2000
Control Group Designs
• Placebo Group (aka negative control)– Alcohol flavored Jello shots– THC free cigarettes
• Wait List Control Group: depression example
Control Example
• Women (N=47) were randomly assigned to:– Subliminal: tapes consistent with labeling– Placebo: tapes inconsistent with labeling– Wait List: received tapes at end of 6 week study
Mirikle & Skanes (1992) J Applied Psychology, 77, 772-776.
Yoked Control• Participant groups matched based on prior trauma
(moderate), sex, and age• Eye Movement Desensitization & Reprocessing: imagine trauma + follow therapists
rapidly moving finger• Control: imagine trauma while staring at stationary object (same duration as
experimental
**
Dunn et al. (1996). J Behav Therapy Exp Psychiatry, 27, 231-239.
Yoked Control
• Part I:– Experimental: no barrier, light than jump– Yoked: barrier, gets same # of shocks
• Part II (no barrier)– Experimental: learned to avoid footshock– Yoked: learned to not avoid footshock
Summaryt-test ANOVA
assumptions Interval/ratio, normal, homogeneity of variance
Interval/ratio, normal, homogeneity of variance
groups 2 3+
statistic t, p < .05 F, p < .05
variations between, within between, within
top related