Experimental design: Between-subject design 8_WinSsdesign.pdf · Between-subject design (Bet-Ss) aka “Between-participant design” Experimental design and vocabulary Threats to

Chapter 9

Between-subject design (Bet-Ss) aka “Between-participant design” Experimental design and vocabulary Threats to internal validity Threats to external validity

Within-subject design (W/in-Ss) aka “Correlated-groups designs” Within-participants design or “Repeated measures design” ▪ Participants complete all (2+) conditions. ▪ Researcher examines if difference between means ▪ e.g. pre-post-test, longitudinal study (vs. cross-sectional) ▪ Multiple testing sessions or complete conditions successively

Matched-participants design

Introduction Can we improve memory for nonsense pics w/ cue?

Method Study: See 28 pics (w/ or w/o label) Test: Immediate recall (draw)

Results Label: 70% (SE= 1.25) No label 51% (SE=.92) t(16) = 3.43, p < .01

Discussion Memory significantly improved by use of verbal label

Droodles by Roger Price a. 4 elephants sniffing an orange b. An early bird catching a very strong worm c. A man in a mailbox signaling a left turn

If Bower, Karlin, & Dueck was replicated using a within-subject design, what would it look like?

What is the independent variable (IV)?

What are the levels of the IV?

What is the dependent variable (DV)?

What is measured?

Bet-Ss desgin W/in-Ss design

Ss Group Accuracy

1 no label 50

2 label 90

3 no label 60

4 label 85

5 no label 45

6 label 75

7 no label 30

8 label 90

Ss Acc-Label Acc-NoLabel

1 90 50

2 85 60

3 75 45

4 90 30

MORE data from fewer Ss!!

05

101520253035404550

Nonsense Rare Common Color-

Associated

Similar

Color-Word

Incongruent

Color-Word

Interference (Seconds)

Co

nd

itio

ns

Inte

rfe

ren

ce

(s

)

Condition

Stroop experiment: “say color of stimulus”

IV: condition (separate groups of Ss)

Duke undergraduates tested day after 9/11 (Flashbulb memory - FB)

Tested again at 1, 6, or 32 weeks after 9/11

DV: consistency of FB and everyday memory

Results: Both types of memory declined over time

No relationship between accuracy and confidence in memory

Method IV: group (training vs. no training) IV: Training - 12+ days on working memory task

“n-back” http://brainworkshop.sourceforge.net/

DV: Performance on intelligence task before and after training sessions

Method: Baseline task:

Find repeated # in series of 10 #s

Transformation task:

Add or subtract 1 to each #, then find repeated # in series of 10 #s

IV: Age (Yng v Older) DV: Accuracy

0.00

0.20

0.40

0.60

0.80

1.00

Baseline TransformationTask conditions

Accu

racy

Yng

Older

PRO

Need fewer Ss (compared to bet-Ss)

Or get more data!

Can take less time If all conditions are in 1 session rather than separate sessions for each Ss

Increases POWER Decreases variability

Less individual differences so more likely effect is due to IV not b/c Ss in grps are different

CON

Testing effects Fatigue effect

Practice effect

Order effect Carry-over effects (“transfer”)

1st condition affects performance in 2nd condition

e.g. knowledge, attitude, etc. Increase in demand

characteristics aka “participant effects”

Ss has more info about the study so can change behavior

Distribute potential order effects evenly over groups by using counterbalancing techniques

If use w/in-Ss design for Bower, Karlin & Dueck ½ Ss receive label then no-label condition ½ Ss receive no label then label condition

Complete counterbalancing If 4 conditions = 24 possible orders!

Partial counterbalancing Randomly select orders to use

Latin squares design Same # orders as # of conditions Each condition presented at each order\ E.g. 3 conditions: ABC / BCA / CAB

Gil Einstein @ Furman University

Event-based vs. time-based PM

Einstein & McDaniel (1990)

Question: What is effect of cue familiarity on PM

Method: Press key when see cue word

Results: 3x more likely to press key for unfamiliar cue word than familiar cue word

Kliegel et al (2004)

How likely are you to forget PM task? - Effect of background task - Effect of importance

Use separate groups of Ss for each condition BUT, Ss for each condition are similar What variables should Ss be matched on? How many variables should Ss be matched on?

It depends on the study!

Used (more often) in clinical or drug research

Match based on:

Person variables (gender, age, ethnicity)

Others such as intelligence, socioeconomic status

Can also use “natural pairs” such as twins

Ss fill-out “test of language ability”

Unknowingly “primed” w/ stereotypes of words in 1 of 3 categories

Examine how quickly Ss interrupts conversation of experimenter w/ friend

PRO

Eliminate testing effects (carry-over or order effects)

Reduce demand characteristics (participant effects)

Less variability among Ss compared to traditional bet-Ss design so increases power

Can use statistics for w/in Ss design so there is more power than bet-Ss design

CON

Need more Ss compared to w/in-Ss design

If loose 1 Ss from one group, loose matched Ss in other group (pair-wise loss)

Need to decide what & # of characteristics to match Ss on

Difficult to find matches (especially if mult. charcts)

Reduces benefit of random selection (can loose benefit of random assignment)

College students rate pronounceability of famous and non-famous male and female names

Later (24hrs/48hrs) judged whether each name was famous or not

50

55

60

65

70

75

80

Familiar M Familiar F Unfamiliar M Unfamiliar F

Per

cen

t hits

Male

Female

74%

65% 68%

61%

0

2

4

6

8

10

Familiar M Familiar F Unfamiliar M Unfamiliar F

Per

cent

fal

se a

larm

s

Nonequivalent control grp Use random assignment Use pretest/posttest design

History Test at different time pts

Maturation Use control group

Testing effect Use control group

Regression to mean Use control grp w/ same extreme scores

Instrumentation effect Use control group

Mortality or attrition Use control group

Diffusion of treatment Tell Ss not to discuss study

Experimenter or participant effects

Use single-blind or double-blind method Use placebo group

Ceiling and floor effects Carefully select DV to avoid

How to prevent these potential confounds