Chapter 8 Experimental Design Dependent Groups and Mixed

Chapter 8 Experimental Design: Dependent Groups and Mixed Groups Designs

Dependent Groups Designs n Matched Designs n Within-Participants Designs – More than one IV – WP Factorial Design

Repeated Measures/With-Participants Design n Each participant is his or her own control. – Increased statistical power (the likelihood of detecting an effect if one is present). – More economical since fewer subjects are needed n Often the DV is assessed at multiple time points such as before and after a “treatment” – Repeated Measures n Repeating treatment conditions enables the participants to identify what is being manipulated. – Demand Characteristics § Concern over demand characteristics prevents more widespread use of repeated measures designs

Repeated Measures/Within. Participants Design n The potential for increased economy and statistical power is weighed against the potential threats to internal validity – Confounds - conditions that vary systematically with changes in the level of the independent variable § When present in an experiment, it is impossible to tell whether changes in the dependent variable resulted from the different levels of the independent variable or from the different levels of the confounded variable. – History – Maturation – Testing

History n Anything that happens between the pretest and posttest that is not part of the experimental situation n The longer the interval between pretest and posttest, the greater the potential for history effect n It is important to know that this is not specifically time passage, but events that occur during that time – Equipment issues – Life events

To Minimize History Effects n Shorten the interval between the pretest and posttest, so things will not happen n Control the environment the pretest and posttest as much as possible

Maturation n Internal processes that occur as a function of the passage of time – Growth and aging processes – Motivational effects such as practice and fatigue n Especially a problem in: – Longitudinal research – Educational research – Therapy research

To Minimize Maturation Effects n Minimize the interval of time between pretest and posttest n Keep all experimental conditions identical during pretest and posttest

Testing n Taking a test once affects scores on the second test – Taking a pretest affects scores on a posttest § Stroop

To Minimize Testing Effects n Use alternate forms, if available n Lengthen the interval between pretest and posttest

Determining the Levels of the Factor n n There should be enough levels to represent the range of values of the treatment variable. There should be enough levels to show the exact nature of the relationship being tested. A true control condition is one in which the treatment variable is absent. Certain problems arise when manipulating the IV – Carry Over Effects

Carry Over Effects n Because each participant experiences multiple treatment combinations, experiencing the early treatments can affect responses to subsequent treatments – Also called “transfer effects” § Order Effects § Differential Order Effects

Dealing with Order Effects n Randomly determine the order of treatments – Difficult to do unless you have a large number of participants. n Completely counterbalanced approach – Requires that each condition occurs equally often, and precedes and follows all other conditions the same number of times. n Incomplete counterbalancing – Requires that each condition occurs equally often.

Counterbalancing Participant 1 2 3 4 5 6 Order of Conditions LA MA HA LA MA LA HA MA MA LA HA HA MA LA

Differential Order Effects n When some levels of the IV may irreversibly influence the DV – Certain orders may permanently change the individuals § Example – A teaching technique n Cannot be controlled by counterbalancing – Need to do a Between-Subjects Design

Repeated Measures Designs ADVANTAGES n Require fewer subjects n Take less time to complete n More powerful than between subjects designs because each participant is compared with him/herself – Error variance is reduced DISADVANTAGES n Problems posed by history, testing, and maturation n Problems posed by multiple treatment interference effects – Differential Order Effects

Example – THC, Alcohol & Driving n Many studies have shown that both THC and alcohol impair driving ability. n No studies have compared the two drugs, nor have any studies examined subjective experiences of the drugs while driving n How do we do this study with a within subjects design?

Starting with the Title

Data Analysis - Partitioning the Variance n The treatment effect is estimated from differences within subjects rather than between subjects. n Between subjects variance reflects differences between subjects, NOT due to the treatment.

Analysis of Variance Summary Table n Between subjects variance is listed first and then removed from further consideration n Within subjects variance is partitioned into the variance due to the treatment and error variance (variance due to chance factors). n The F ratio compares the variance due to the treatment (numerator) to the error variance (denominator).

Interpretation of the Results of ANOVA n Calculated F is compared to critical F: If calculated f is equal to or greater than critical F: n F is significant n There is a significant difference among the means of the different treatment levels

Get it?

Data Analysis

Results

Discussion n n Describes the outcome of the research in words – Briefly summarize what you found Integrates the outcome of the study with previous research findings – Here is how you data fit in to the larger literature base Draws conclusions – Based on theory or practical application May present suggestions for future research – Limitations

Discussion – Explaining the effect

Discussion – Implications & Meaning

Discussion – Addressing a limitation

Discussion – A summary of the results that happens to be a conclusion

Discussion – Addressing a limitation

Example – 2 X 2 WS Factorial

Question/Problem n Increase in the coadministration of alcohol and caffeine (energy drinks) in college students n Little research on the interactive effects. n How does caffeine influence the effects of alcohol on information processing tasks?

Methods

Procedures & Results

Mixed Designs n Have at least two independent variables n At least one variable is a between subjects factor. n At least one variable is a within subjects factor.

Reasons for Using Mixed Designs n Repeated measures factors are desirable because they require fewer participants and can take less time. n Some variables are manipulated between participants to reduce or prevent fatigue, interference effects, or demand characteristics. n Sometimes participant variables are studied (e. g. , gender, handedness, etc. )

Example - Our THC and STM Study n n Problem – The effects of THC Intoxication on the ability to do occupational tasks requiring STM Research Hypothesis – THC intoxication will impair STM IV – Three smoked THC doses – 0%, 5%, 10% DV – Span test for words at different time intervals – 15 min, 1 hr and 3 hrs

Descriptives

Testing the WP Factor

Testing Levels of WP Factor

Testing the BS Factor

Post Hoc Test on the BS Factor

Matched groups designs n Provides the added power of dependent groups designs and eliminates the problem of carryover effects. n Matching variables – participants are matched, or made equivalent, on variables that correlate with the DV. n By using appropriate matching variables the error variance due to individual differences is reduced and power is increased.