Psych 250: Exam 3: Ch. 11
What does experimental research typically equal? But in the real world?
1 IV & 1 DV But in the real world, variables rarely exist in isolation Meaning, relationships are more complex
A researcher is interested in examining the effects of mood and food deprivation on eating. Female participants listen to one of 2 types of music to induce either a happy or a sad mood, following either 19 hours of food deprivation (breakfast and lunch skipped) or no food deprivation. The participants are then given free access to food in a controlled laboratory setting, and the amount of food consumed is measured for each individual. How many factors? What are the factors? How would you describe this study using the notation system?
2 Mood & food deprivation 2(mood: sad or happy) X 2(food deprivation: deprived or no deprivation)
How Many Factors/Levels for: 2 X 2 factorial design? 2 X 3 X 2 factorial design?
2 Factors, 4 Levels 3 Factors, 12 Levels
How was Liguori and Robinson's design setup?
2 IVs/factors: Alcohol (2 levels) & Caffeine (3 levels) So 2 X 3 factorial design = 6 total treatment conditions DV = Reaction Time (how quickly participants applied breaks in a simulated driving test)
What does Factorial mixed designs combine? Give an example.
2 different research designs (commonly 1 between-subjects factor & 1 within-subjects factor) Ex. 1 factor expected to produce large order effects
What is Higher-Order Factorial Designs?
3+ factors (Ex. 3 X 3 X 3 factorial design) Same basic concepts of 2-factor designs apply, now you just have more factors
In Factorial Combined Strategy Designs, Combined Strategies=? Give an example.
= 1 factor manipulated & 1 is not Ex. alcohol & gender 1 IV (alcohol) & 1 quasi-IV Quasi-IV typically falls into 1 of 2 categories: pre-existing characteristic (AKA person-by-environment or person-by-situation designs) time ("mixed/combined design")
In Factorial Combined Strategy Designs, experimental=?
= both factors manipulated Ex. alcohol & caffeine
In Factorial Combined Strategy Designs, NE/QE=? Give an example.
= both factors pre-existing/not manipulated Ex. College math vs no college math, recall at 3 & 5 years Difference between NE & QE = degree confounding variables controlled (threats to internal validity)
No interaction=
=factors independent Meaning the effect of 1 factor on the DV does not depend on the level of the other factor
What is a factorial design?
A research study involving 2+ factors. Often referred to by # of factors Ex. 2 IVs = 2-factor design
What is a single-factor design?
A research study with only 1 IV
What are advantages of between-subjects factorial designs?
Avoids order/carryover/testing effects Best for situations when a lot of participants available, individual differences small, and order effects are likely
Means may not be representative of individual effects?
Because each main effect is an average, it may not accurately represent any of the individual effects used to compute the average
All designs discussed thus far have been:
Between-subjects (i.e., different participants in each condition) True independent variables (i.e., true experiments in which you manipulate variables)
How to expand and replicate in regards researchers to observe influence of 2+ variables acting & interacting simultaneously?
Build on previous research Same results if conditions changed? Administer treatments (1 factor) under several different conditions (2nd factor). Results interact with condition? Same results if participant characteristics changed? Administer treatments (1 factor) to different types of individuals(2nd factor). Results interact with participant type?
What are advantages of within subjects factorial designs?
Can be time-consuming because all participants have to go through each condition (increases attrition) In a factorial design counterbalancing may be difficult, testing effects possible
What are disadvantages of between-subjects factorial designs?
Can require a large sample (Ex. 2 X 4 factorial design = 8 conditions) If you need 30 participants per condition = 240 participants Individual differences between groups can = confounds & increase variance
What does interaction between factors reveal? Give an example.
Combination of 2 factors operating together to affect DV Ex. combined effect of alcohol & caffeine not same as alcohol alone Effect of alcohol on RT depends on amount of caffeine consumed (or vice versa)
How to identify interactions?
Compare mean differences between cells with the mean diffs predicted by main effects
What is multiple regression?
Correlational relationships between several predictor variables & 1 criterion variable
What are between-subjects factorial designs?
Different group of participants for each condition (cell)
Again, main advantage of factorial designs is that you can see:
Effects of each individual factor Effects of the combinations of different levels of factors
Give an example o f3-way interactions: 2-way interaction (Ex. A X B) depends on levels of 3rd factor (C), (Higher-Order Factorial Designs).
Ex. DV = academic performance, & 3 IVs Factor A: 2 teaching methods (2 levels: method 1 & method 2) Factor B: Student gender (2 levels: boys & girls) Factor C: Grade (2 levels: 1st & 2nd grade) 2 X 2 X 2 Higher-Order Factorial Design 3 possible main effects (A, B, C) 3 possible 2-way interactions (A X B, B X C, A X C) 1 possible 3-way interaction (A X B X C) 3-way interactions: 2-way interaction (Ex. A X B) depends on levels of 3rd factor (C)
Give an example of multiple factors at play?
Ex. GPA related to IQ: motivation, parental influence, SES, etc.
Give an example of main effect.
Ex. Main Effect of alcohol: diff between alcohol vs. no alcohol means Main effect of caffeine = diff between caffeine conditions Differences here or here = main effects Main effect of caffeine cannot be explained by alcohol (alcohol matched across conditions, ½ of participants in each caffeine column had alcohol, ½ had no alcohol) Means appear to show faster response times as caffeine increases Main effect of alcohol cannot be explained by caffeine (matched across conditions, 1/3 of participants in alcohol condition had no caffeine, 1/3 had 200mg, 1/3 had 400mg). Means appear to show slower response times with alcohol than with no alcohol
Can blend several different factors, designs, & research strategies we have discussed previously into 1 factorial study
Experimental , or NE/QE Between-subjects or within-subjects
What are not manipulate in Non-experimental and Quasi-experimental? Give an example.
Factors Ex. Bahrick and Hall (1991): examined permanence of memory for high school algebra & geometry Factor 1 = college math courses: Group 1 - those who had taken college math courses Group 2 - those who had not taken college math courses Factor 2 = time Reassessed participants at multiple intervals over time From 3 years up to 55 years after high school graduation
When a study includes 2 or more IV's, IV's are called? Usually denoted as? Notation system identifies? Give an example.
Factors: the IV's in a study with 2+ IV's Usually denoted by letters (A, B, C, etc.) Notation system identifies # of factors (IVs) & # of levels for each factor Ex. In a study with 2 factors (IVs), 1 IV has 2 levels, 1 IV has 3 levels The design would be 2 X 3 factorial design (stated 2 "by" 3) The total # of treatment conditions determined by multiplying 2 X 3 design would have 6 conditions
What is the goal in expanding and replicating in regards to researchers to observe influence of 2+ variables acting & interacting simultaneously?
Goal: Replicate past work AND expand on it in 1 study Replication = repeating previous study by using the same factor/IV exactly how it was used Expansion = adding a 2nd factor in the form of new conditions or participant characteristics Can previous results/effects be generalized to new situations and/or new populations?
Give an example of expanding and replicating in regards to researchers to observe influence of 2+ variables acting & interacting simultaneously?
Green et al. (1994) Ex. Examining perceived value of future rewards Typical study: participants asked to choose between: $1000 payment they will receive in 5 years OR smaller payment today Typical result: the longer the $1000 payment is delayed, the smaller the amount people will accept in exchange for today i.e., person may settle for $500 today if wont get $1000 for 5 years But same person will take $200 today, if $1000 wont come for 10 years Ex. Original studies were conducted using only college student samples Green et al. replicated results with college students AND expanded the previous research by adding participant age as a new IV Results: interaction between age and delay of payment Effect of increasing the delay period depended on the age of the participants Younger people show a quick drop-off in value of future reward as delay increased Drop-off for older people was more gradual
What do main effects reveal? What does this mean?
How each factor affects DV i.e., Results that would be obtained if each factor were examined in separate experiments
What was was Liguori and Robinson's methods?
IV/factor = Alcohol (2 levels): no alcohol or 0.6 grams ethanol per kilogram of body weight (i.e., approx. 8 ounces of wine per 100 lbs for each participant) IV/factor = Caffeine (3 levels): no caffeine, 200mg, or 400 mg DV = Reaction time 2(Alcohol) X 3(Caffeine) design = 6 conditions
Give an example of the advantages of using a factorial design (regarding the alcohol & caffeine study)?
If we did 1 study examining effects of alcohol, we would only be able to observe the effects of alcohol on reaction time If we did 1 study examining the effects of caffeine, we would only be able to observe the effects of caffeine on reaction time A factorial design permits explanation of how changes in caffeine consumption can influence effects of alcohol on reaction time
What does factorial designs include?
Include more than 1 IV Useful for NE, QE, & Experimental strategies
How to reduce variance between groups in regards researchers to observe influence of 2+ variables acting & interacting simultaneously?
Individual differences may = problems for between-subjects designs via large variance within conditions You can remove the variable (hold constant/reduce range) but doing this limits external validity Solution: Include variable contributing to high variance (Ex. age) as factor Groups now divided into age factor levels (Ex. young vs. older)
What should you allow researchers to observe? And discuss 3 specific situations it's high applicable.
Influence of 2+ variables acting & interacting simultaneously Highly applicable to 3 specific situations: Expanding & replicating previous work Reducing variance in between-subjects designs Evaluating order effects in within-subjects designs
Interaction=
Interaction=mean differences between cells are not explained by main effects
Factorial designs are represented by?
Levels of 1 factor form the rows Levels of other factor form the columns
Who addressed this question: If someone has been drinking alcohol, will coffee (or caffeine) really improve their reaction time while driving?
Liguori and Robinson (2001)
Main effects could be distorted?
Main effect for 1 factor is obtained by averaging all the different levels of the 2nd factor, thus main effects will be distorted if there is an interaction
What is Symmetrical order efects in regards to evaluate order effects within groups?
Main effect of treatment & order, interaction Effect of 1 factor (treatment) depends on the other (order) Symmetrical - the treatment that comes 2nd is affected the same regardless of whether it's A or B
What is non symmetrical order effects in regards to evaluate order effects within groups?
Main effect of treatment & order, interaction (nonsymmetrical graph) Effect of 1 factor (treatment) depends on the other (order) Nonsymmetrical - order effects of treatment A are not as great as order effects of treatment B
What is No order effects in regards to evaluate order effects within groups?
Main effect of treatment, but not order, no interaction i.e. doesn't matter which treatment comes first
Any combination of factors
Mixed designs Combined designs
What is the advantage of using a factorial design?
More realistic situation (increases external validity)
What are experimental factorial designs?
More than 1 IV (called factors) Examines how different variables interact 2 X 3 design = 2 levels of factor A & 3 levels of factor B Ex. At dinner parties, common for host to serve coffee at the end if guests have drank alcohol during the evening so that guests are more alert for the drive home. But does this really work?
Two terms that are associated with researching multiple independent variables?
Multiple Regression Factorial Designs
What are 3 possible outcomes when examining order effects by using order as a between-subjects factor?
No order effects Symmetrical order effects Non symmetrical order effects
What is another way of defining interactions? Give an example.
Notion of independence between factors No interaction = main effects are "independent" (factors do not influence each other) Interaction = factors are "interdependent" (1 factor influences effect of the other) Ex. Size of alcohol effect (top vs. bottom row) depends on level of caffeine
How to evaluate order effects within groups in regards researchers to observe influence of 2+ variables acting & interacting simultaneously?
Order effects = serious problem for within-subjects designs Sometimes a researcher may be interested in how order effects DV? Ex. Is treatment A more effective if it comes after treatment B, or before treatment B? Add order as between-subjects factor (i.e., mixed design) Treatment is the within-subjects factor Order is the between-subjects Now you can examine nature & magnitude of order effects
What is Factorial Pretest-Posttest Control Group Designs?
Regular Quasi-Experimental design (AKA 2-factor mixed design) Participants cant be randomly assigned to 1 group or another Factor 1 - treatment or control (cant be assigned, ex. Depressed or not) between-subjects quasi-experimental variable Factor 2 - pre or post treatment observation, within-subjects variable O X O (treatment group) O O (nonequivalent control group) IF researcher can use random assignment to 1 of 2 groups for an IV = combined experimental & non/quasi-experimental strategy 1 IV (experimental): treatment/control (participants randomly assigned) 1 quasi-IV (pre-post): within-subjects variable R O X O (treatment group) R O O (equivalent control group) O=observation/measurement X=treatment R=random assignment
What are disadvantages of within subjects factorial designs?
Requires fewer participants Individual differences less likely to be confounding variables
What is the biggest advantage of using a factorial design? Why not just do 2 studies, 1 testing each variable?
See how individual factors alone & in combination with each other influence DV. Because then we couldn't see how the combination of the factors influences the DV
What are within subjects factorial designs?
Single group for all conditions
What is the definition of interaction?
Unique mean differences in DV that are not explained by the main effects of factors/IVs
What is the interaction between factors (interaction effect)? Give an example.
When 1 factor has a direct influence on effect of a second factor on the DV i.e., whenever 2 factors, acting together, produce mean differences that are not explained by the main effects of the two factors Ex. If effect of alcohol on RT depends on amount of caffeine consumed Main effect of alcohol does not apply equally across caffeine conditions
Main effect for 1 factor obtained by averaging all levels of other factor
With an interaction the factors are interdependent Meaning the effects of factor A depend on levels of factor B Thus, generally speaking: If there is a significant interaction you ignore main effects because they are likely to be distorted
Slide 32 picture review
Within-subjects factor: All see list of pleasant & unpleasant words, recall tested. Between-subjects factor: 2 groups 1 group listens to happy music & other sad music.
What are mixed designs?
both Between-subjects & within-subjects
Each cell =
each condition (i.e., specific combination of the levels of each factor)
What are combined designs?
experimental & NE/QE elements
Mean difference between conditions for single factor =
main effect
Significant interaction =
main effects of factors/IVs are not straightforward
Be cautious with main effects if you have a
significant interaction
