1 / 51

Psychology 242, Dr. McKirnan

Psychology 242, Dr. McKirnan. Complex experiments: Within- subjects & blocking designs. Own control Reversal designs Repeated measures & Randomized block designs. Final Exam: Wednesday the 7 th 1:00 to 3:00, in the class room. 11/22/11. Basic forms of within-subjects designs, 1.

zinna
Télécharger la présentation

Psychology 242, Dr. McKirnan

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Psychology 242, Dr. McKirnan Complex experiments: Within- subjects & blocking designs • Own control • Reversal designs • Repeated measures & Randomized block designs Final Exam: Wednesday the 7th 1:00 to 3:00, in the class room. 11/22/11 Within Subjects Designs

  2. Basic forms of within-subjects designs, 1 Basic forms of within subjects designs; • Own control • Each participant in control and experimental group. • Optimally, order is counter-balanced • Reversal designs • Repeated measures & Randomized block designs Within Subjects Designs

  3. Basic forms of within-subjects designs, 3 Basic forms of Within subjects designs; • Own control • Reversal designs • Hypothesis: behavior controlled by clearly bounded condition • Design: “A – B – A”; impose – withdraw – impose condition • Repeated measures & Randomized block designs Within Subjects Designs

  4. Basic forms of within-subjects designs, 2 Basic forms of Within subjects designs; • Own control • Reversal designs • Repeated measures • Multiple treatment conditions:each participant gets each treatment. • Longitudinal / time sampling:each participant assessed over multiple time periods • Randomized block designs; Repeated measure combined with between-groups variable. Within Subjects Designs

  5. Psychology 242, Dr. McKirnan Complex experiments • “Own Control” designs • Each participant is in the control and experimental group. • Optimally, order is counter-balanced • Reversal designs • Repeated measures & Randomized block designs  Within Subjects Designs

  6. What do you hear? Write down your answer: A = ‘aa’ B = ‘ba’ C = ‘ca’ D = ‘da’ E = ‘ga’ Within Subjects Designs

  7. What do you see? Write down your answer: A = ‘aa’ B = ‘ba’ C = ‘ca’ D = ‘da’ E = ‘ga’ Within Subjects Designs

  8. What do you hear? Write down your answer: A = ‘aa’ B = ‘ba’ C = ‘ca’ D = ‘da’ E = ‘ga’ Within Subjects Designs

  9. This experiment • The video presents the syllable “ga” • The audio presents “ba” • The brain “fuses” them so most people perceive “da”. • This is an audio-visual illusion. • The within-subjects design allows us to clearly contrast these within the same people. Within Subjects Designs

  10. Observe1 Observe2 Between-subjects designs Basic experiments: “Between-Subjects” designs Group 1 Experimental condition /Treatment Group 2 Control condition 2 separate groups: Observed (naturally occurring)or RandomlyAssigned to be equivalent Independent Variable: One group receives the experimental condition or treatment, one does not. Dependent Variable(s): Measured in both groups. • Clear experimental manipulation; treatment given to only one group. • Hypothesis tested by differences between groups. • Internal validity:groups must differ only on Independent variable; non-equivalence at baseline = confound • Statistical power requires large number of subjects. Within Subjects Designs

  11. Why do “Within – Subjects” designs? Within Ss designs • Many research questions address contrasts between different states within one person • Alcohol v. non-alcohol use  aggression, risk • Learning condition v. recall state  “State dependent learning” • Many studies address change over time • Behavioral or biomedical intervention studies • “Natural history” or cohort studies • Practical efficiency of within Ss designs • Powerful contrasts within a participant; less within-group (error) variance. • Studies that require rare or costly participants. Within Subjects Designs

  12. Pre- Post- designs Non-experimental within-subjects approach Single Group Experimental Manipulation Baseline (Observe1) Follow-up (Observe2) All participants are assessed at baseline All participants then get the Experimentalintervention and follow-upmeasurement. Single Group 10-week smoking cessation program Assess smoking rate Follow-up smoking rate Within Subjects Designs

  13. Pre- Post- designs Single Group Experimental Manipulation Baseline (Observe1) Follow-up (Observe2) All participants are assessed at baseline All participants then get the Experimentalintervention and follow-upmeasurement. • Hypothesis is tested by change from baseline • This is a main effect analysis • “Groups” = baseline v. follow-up scores • Possible confounds? • History • Maturation • Regression… Within Subjects Designs

  14. Observe1 Experimental Condition Observe2 Control Condition “Own Control” True experiment with within-subjects design Single Group All participants get the Control condition and measurement All participants then get the Experimentalintervention and measurement. Single Group Recall task Learning under ‘white noise’ condition Recall task Learning under quiet conditions Within Subjects Designs

  15. Observe1 Experimental Condition Observe2 Control Condition “Own Control” Single Group • Each participant is his own “control group” • Hypothesistested by differences between conditions(Observation1 v. Observation2)within group. • Internal validity:eliminate possible confound of group differences at baseline, since there is only one group. • Experimental manipulation potentially less clear due to possible carry-over effects. • Statistical powerincreased: requires fewer subjects. Within Subjects Designs

  16. Own Control design with counter-balancing Group 1 Observe1 Experimental Condition Observe2 Control Condition Group 2 Observe1 Control Condition Observe2 Experimental Condition 2 groups, naturally occurring orrandomly assigned Basic own-control design done twice. Group 1 Recall task Learning under ‘white noise’ condition Recall task Learning under quiet conditions Group 2 Learning under ‘white noise’ condition Recall task Learning under quiet conditions Recall task Within Subjects Designs

  17. Own Control design with counter-balancing Group 1 Observe1 Experimental Condition Observe2 Control Condition Group 2 Observe1 Control Condition Observe2 Experimental Condition Own-control design done twice. • Test the Hypothesis: by combining the 2 groups. • Internal validity: • eliminate confound of group differences at baseline. • Lessens (& allows for test of) carry-over effects • Statistical power design requires more subjects. Within Subjects Designs

  18. Survey example of own-control design Phenomenon: Treatments for HIV lower the amount of virus in the blood (“viral load”), which may make the person less infectious. Hypotheses: 1. Gay men decide how risky an HIV-infected partner is based on whether the partner has a low viral load. 2. Men who are particularly risk-prone are more likely to decide someone is “safe” via viral load information. Vanable, P.V., Ostrow, D.G., McKirnan, D.J., Tayawaditep, J., & Hope, B.A. (2000). Impact of combination therapies on HIV risk perceptions and sexual risk among HIV-positive and HIV-negative gay and bisexual men. Health Psychology,19(2), 1-12. Within Subjects Designs

  19. Design: Community survey of Chicago MSMs IV# 1; Repeated measure: hypothetical partner • – The participant reads about each of 2 potential partners: • An HIV+ man who is not in treatment • An HIV+ men in treatment, with a low viral load • – Then rates how risky each partner might be. True independent variable; manipulated by the experimenter. IV #2; Blocking variable: participants’ risk status – Participants describe their history of sexual risk and are categorized by the experimenter as ‘high’ v. ‘low’ risk Measured “blocking” variable only. Within Subjects Designs

  20. Mixed Repeated Measures: data structure Repeated measure: All participants respond to both conditions. Blocking variable: Men “assigned” to high v. low risk groups based on their interview answers. 2, 3, 11, 5, 9, 12, 13, 16…n = 488 2, 3, 11, 5, 9, 12, 13, 16…n = 488 1, 4, 6, 7, 8, 10, 14, 15…n = 66 1, 4, 6, 7, 8, 10, 14, 15…n = 66 Within Subjects Designs

  21. Survey: repeated & blocked analyses, 4 Data:A statistical interaction between the repeated measure and the blocking variable. Vanable, P.V., Ostrow, D.G., McKirnan, D.J., Tayawaditep, J., & Hope, B.A. (2000). Impact of combination therapies on HIV risk perceptions and sexual risk among HIV-positive and HIV-negative gay and bisexual men. Health Psychology,19(2), 1-12. Within Subjects Designs

  22. Survey: repeated & blocked analyses, 5 Data: Both risk groups consider an HIV+ partner with low viral load to be safer “Low” viral load information lowers risk perception most for the risky men. Giving all participants both conditions creates a strong contrast (Independent Var.). Within Subjects Designs

  23. Advantages of own-control designs • Participant assignment problems eliminated, since there is only one group. • No need for random assignment or matching procedures • No possible problem with groups differing at baseline. • Can use a smaller n than between S’s • Half as many participants for simple own-control experiment • Very useful for rare or expensive participants • More sensitive to experimental effects; less within-subject (error) variance Within Subjects Designs

  24. Disadvantages of own-control designs • Sequencing or order effects: • Scores change if participants learn or get practice over different conditions • Participants get sensitized to experiment or experimental procedures • Cure by counter-balancing order of condition. • Takes as many participants as between - group design • Participant “burden” • Fatigue from multiple experimental tasks • Possible drop-out Within Subjects Designs

  25. Psychology 242, Dr. McKirnan Complex experiments • Own control • Reversal designs • Hypothesis: behavior is controlled by a clearly bounded condition • Design: “A – B – A”; impose – withdraw – impose condition • Repeated measures & Randomized block designs  Within Subjects Designs

  26. Basic Reversal design structure Impose temporary experimental condition “Normal” baseline state Return to normal state Measurement or testing Measurement or testing Measurement or testing • Examples: • Role of incentives in enhancing performance • Impact of anti-depressant drug on mood • Effect of self-awareness on following social norms Within Subjects Designs

  27. Core Assumptions of reversal designs: Clear stimulus or condition -related hypothesis • Hypothesize that behavior is directly tied to a condition or stimulus • Clear beginning and ending • No “carry-over” effects (due to, e.g., learning, sensitization, etc.) • Changes in behavior only last as long as condition is in place • Changes can be induced and reversed more than once Within Subjects Designs

  28. Participant’s drinking rate Examples of reversal designs Test effect of, e.g., modeling (observation of attractive experimental confederate) on alcohol consumption. • If the model influences participant’s behavior: • Consumption will increase when the model’s does… • Rate goes back down when model’s does. • Up again with model, etc.. Within Subjects Designs

  29. Reversal designs & carry-over Reversal designs can test carry-over effects: Carry-over effect: Drinking rate gradually increases over time no matter what. Reversal effect: Modeling controls drinking rate Within Subjects Designs

  30. Example: single group study with reversal design The Hawthorne Study. • The Hawthorne Study is one of the more… • Famous, • Corrupt, • Misunderstood & misused. • social research studies ever conducted • Context: • Hawthorne Electrical Plant (Hawthorne Il.) • 1950s era • Strong political fears of “Communist conspiracy” • Unions seen as ‘fronts’ for international Communism. • Working conditions terrible at Hawthorne, as in many companies. • OSHA did not exist; worker’s rights largely absent • Many Social Psychologists were imbued with the conservative (anti-union) tenor of the times.

  31. Single group reversal design: The Hawthorne Study • Context(cont.): • Union drive taking place • Demands for better working conditions • Productivity & quality decreasing Pre-Intervention Baseline assessment 1st Intervention (change lighting) Study Purpose • Increase motivation & productivity. 1st Follow-up assessment Hypothesis • Employees simply want attention • Any change in the work environment - even a trivial one - leads to change. 2nd Intervention (Reversal; change the lighting back) Intervention • Change inadequate lighting. 2nd Follow-up assessment Study structure: Reversal design Week 12-13, quasi-experimental designs.

  32. Single group reversal design: The Hawthorne Study • Outcomes • Daily assembly line output Findings • Output rises after lighting improved (1st follow-up). • Output rises again after lighting reversed to initial state. • Thus: • Any change, even a negative one, “motivates” workers.. • Workers respond to simple attention, not real change. Internal validity? • Political bias: researchers hired to disprove workers’ claims! • Mortality: as part of “union busting” dissatisfied workers fired during study period. • Reactive measures: workers fear for job may increase production, not workplace change. Sad legacy of the Hawthorne study • The “Hawthorne effect” is commonly cited to discount demands for change, or explain away positive findings of interventions.

  33. Psychology 242, Dr. McKirnan Complex experiments • “Own Control” designs • Reversal designs • Repeated measures & Randomized block designs • Combine a blocking or grouping variable with a repeated measure. • Most common Within-Subjects design • Biomedical research • Behavioral intervention evaluation 

  34. Simple repeated measures / time series designs Group Group3 Group2 Measure1 Measure1 Measure1 M2 M2 M2 M3 M3 M3 M4 M4 M4 M5 M5 M5 M6… M6… M6… • Examine / describe changes over time in one or more key variables. • Describe or test hypotheses about group differences over time. • Groups may be assigned, in a true experiment. • … intervention groups with long-term follow-up • Groups may be measured or naturally occurring. • … age, gender or ethnic groups. • Longer time-frame yields more valid & interpretable data.

  35. Isattention to childhood obesity causing it to decease? • 2003  2012 data • Older kids (2 - 19): no change • Toddlers appear to be doing better. • Supports effectiveness of recent infant programs. E X A M P L E • Longer time frame: 1999  2012 • Older kids no still show no change • Toddlers only look better because of a spike in 2003. • Looking back to 1999 shows a flat line with lots of variance.

  36. Interrupted time series design Group Measure1 M2 M3 M4 M5 M6… Intervention or event • Test effect of intervention or event on ongoing series of measurements. • Intervention may be experimental or observed • Policy shift, e.g., educational policy • Uncontrolled event; e.g., 9/11/01, Media event • Assessments may be experimental or archival • Successive cross-sectional surveys • Traffic data, clinic or crime reports, test scores

  37. Time series designs Group Measure1 M2 M3 M4 M5 M6… Intervention or event • Multiple baseline • Demonstrate highly stable effect • long-term crime rates • disease prevalence • economic performance… • Show steady rate of change • Hypothesis; tested by: • Shift in stable rate after intervention • Increase / decrease in rate of change after intervention

  38. Example of interrupted time series:Shift in Baboon culture. Core question: Do baboon troops develop and transmit a learned “culture”? Baseline: Long-term observational data on aggressiveness in a specific baboon troop. • Intervention: • Tuberculosis outbreak due to infected food. • Dominant / aggressive males fed first • are selectively infected • are naturally culled from troop • Naturally occurring event in >20yr. ongoing field study: Long repeated measures / time series. E X A M P L E

  39. Baboon culture: findings Quasi-controls: Parallel data from other baboon troops. • Core finding: • With dominant males gone, remaining males showed more cooperative behavior • Enhanced cooperation was transmitted across generation, showing learned “culture”. • Major virtue of repeated measures: • Finding made possible by many assessments over considerable time E X A M P L E

  40. Randomized block designs Blocking Variable;between - subjects factor • “Person” variable;age, gender, ethnicity, etc. • Not a “true” IV since people not randomly assigned; Or: • Experimental condition; drug dose, treatment, etc. • “True” IV with random assignment Repeated measure: within-subjects factor • Multiple treatment conditions: each participant is observed in each treatment condition • (e.g., high v. low drug dose, different instructions…) Or: • Longitudinal / time sampling: measure D.V. over multiple time periods (Cohort studies)

  41. Baseline Measure Measure2 M3 M4.. Measure2 M3 M4.. Baseline Measure Within subjects designs; own control, 3 Repeated measures / randomized block design Group 1 Control Condition Group 2 Experimental Condition Assignment Randomly or via natural “blocks” Treatment. Primary Independent Variable. Control group may receive Placebo. Baseline assessment prior to intervention or experimental condition. Follow-up. Long-term assessment of outcome or Dependent Variable. Time may represent 2nd Independent Variable. Within Subjects Designs

  42. Example of Repeated Measure Design Question: Do Gay / bisexual men who use HIV medications as Post-Exposure Prophylaxis [PEP] have more risky sex. • Independent variables: • Medication use:Between subjects / Measured “blocking” variable: men who request PEP v. those who do not. • Study visit [time]: Repeated Measure [within subjects]: Participants are interviewed every 6 months. • Dependent variable: % of men reporting risky sex at any given time period. Within Subjects Designs

  43. Blocking variable Repeated Measure Unprotected anal intercourse (UAI) with HIV positive and unknown sero-status partners among MSM, by study visit Blocking variable Participants reporting (%) All men get safer over time (a Main Effect) OR*(CI)p PEP 1.7(1.2-2.2) .001 Visit 0.97( .96-.99) .001 Month of study visit *Adjusted for age, study site, drug use, and education Within Subjects Designs

  44. Unprotected anal intercourse (UAI) with HIV positive and unknown sero-status partners among MSM, by study visit Blocking variable Participants reporting (%) Non-users of PEP are generally safer (also a Main Effect) OR*(CI)p PEP 1.7(1.2-2.2) .001 Visit 0.97( .96-.99) .001 Month of study visit *Adjusted for age, study site, drug use, and education Within Subjects Designs

  45. Unprotected anal intercourse (UAI) with HIV positive and unknown sero-status partners among MSM, by study visit Blocking variable Participants reporting (%) Non-users at the end of the study are safest (an Additive Effect; time + group) OR*(CI)p PEP 1.7(1.2-2.2) .001 Visit 0.97( .96-.99) .001 Month of study visit *Adjusted for age, study site, drug use, and education Within Subjects Designs

  46. Unprotected anal intercourse (UAI) with HIV positive and unknown sero-status partners among MSM, by study visit Blocking variable Participants reporting (%) There is no Interaction Effect: The effect of time is the same for both groups. OR*(CI)p PEP 1.7(1.2-2.2) .001 Visit 0.97( .96-.99) .001 Month of study visit *Adjusted for age, study site, drug use, and education Within Subjects Designs

  47. Hypothetical Example of a randomized block design; basic clinical intervention / drug trial Hypothetical randomized block design, 1 Question: Effectiveness of different doses of an anti-hypertensive drug 2nd questions: Effects of: time, gender, medical status of patients. Population: Hypertensive patients [systolic Bp > 145] Outcome [D.V.]Systolic Blood pressure Blocking variable: Drug dose: placebo v. 2 doses (high / low). This IV carries the main hypothesis. Repeated Measure:Time. Key element of effectiveness: stability x follow-up. 2nd Blocking vars.: Gender, medical status, ethnicity, etc. Within Subjects Designs

  48. Hypothetical randomized block design, 3 Example # 1 (hypothetical data): drug dose & time on systolic Bp, complete sample • What does this pattern of data show? • Main effects? • Interaction effects? • Hypothesis supported? Systolic blood pressure Within Subjects Designs

  49. Hypothetical randomized block design, 4 Example # 2, hypothetical data: drug dose & time on systolic Bp, complete sample • What does this pattern of data show? • Main effects? • Interactions? • Hypothesis supported? Systolic blood pressure Within Subjects Designs

  50. Hypothetical randomized block design, 5 Example # 3, hypothetical data: drug dose & time on systolic Bp, complete sample • Main effects? • Interactions? • Hypothesis? Systolic blood pressure Within Subjects Designs

More Related