lecture 4 feb 6

Question 1

Describe a Between-Subjects design

Answer 1

• Different subjects within each level of the IV
• Easier to control for time/order effects
• Harder to control for individual differences
• Necessary when subjects must be naïve, or when using a subject variable (cultural or sex differences)
  • E.g., deception experiments, cultural studies
• We must still do our best to avoid confounds
  For this we use some combination of random assignment & matching when creating groups

Question 2

Describe a Within-Subjects design

Answer 2

• Same subjects within each level of the IV
• Easier to control for individual differences
• Harder to control for time/order effects (participants will see every level)
• Every subject participates in every condition
• Full control of extraneous participant variables
  • By ‘removing’ participant variability we can get a cleaner look at our independent variable
• Must be careful about order effects, because performance can change during the experiment

Question 3

Describe Random Assignment

Answer 3

○ Just as we randomly sampled our subjects, we randomly assign them to groups

Question 4

Describe Matching

Answer 4

○ Used to manually control a potential confound
○ Remember, extraneous variables are confounds when they systematically affect your DV
○ Perfect matching can be tedious or impossible
○ Your matching strategy could be as simple as splitting your sample in half
○ E.g., low and high anxiety, with equal numbers of low and high in each group
§ Always consider the reliability & validity of the measurement used for your matching variable
○ Remember: it is impossible to control everything, so we must rely on randomness for the rest. Randomness might be insufficient the smaller your sample is, so you might have to use matching.

Question 5

Describe Progressive effects

Answer 5

§ Practice effects: participants perform better on later conditions than earlier conditions
§ Fatigue effects: participants perform worse on later conditions than earlier conditions

Question 6

Describe Context effects

Answer 6

participants change how they perceive or approach the experiment
§ These effects can be difficult to anticipate

Question 7

Describe Complete counterbalancing

Answer 7

○ Make equal the number of participants who complete the experiment in each possible order
1. It allows us to control the effect of the order
2. Allows us to measure the effect of the order
○ Downside: if you have many levels, the # of possible orders can get impractically large
§ # conditions = #!

Question 8

Describe Partial counterbalancing

Answer 8

○ If you have too many levels, you can randomly choose an order for each participant
○ However, like all random procedures, if your sample is small, biases can occur
○ A condition may be affected by:
§ Its position in the experiment
§ The condition immediately before or after it

Question 9

Describe The Latin Square

Answer 9

○ Controls both position and what comes immediately before & after

Question 10

Describe usefulness of Many Measurements

Answer 10

• Random ordering is more effective in this case
  
  • To obtain more control, we can use blocking
    Reverse counterbalancing

Question 11

Describe Cross-sectional designs

Answer 11

• E.g., designing a study to test whether memory capacity declines with age
  
  • Can split age into 5 year groups:
    ○ 10-20, 20-30, 30-40, 40-50, 50-60, 70-80
  • What have we failed to control here?
    These are called cohort effects

Question 12

Describe Longitudinal Design

Answer 12

• E.g., designing a study to test whether memory capacity declines with age
  
  • Recruit a sample of 20 year-olds and measure them every year (or every 2 years, 5 years, etc.)
  • Because this measure is within-subjects, all extraneous participant variables are controlled

Question 13

Describe Confounds in longitudinal designs

Answer 13

• Attrition effects: participants will drop out during the experiment. Some participants may be more likely to drop out, creating a confound.
  
  • Practice effects: participants can improve on your measure for reasons other than age
  • Instrumentation effects: the way you take your measurements may change over the years
  • Cohort effects: because all your participants were recruited at the same time, they make up one cohort, with potentially unique aging effects
  • To help control cohort and attrition effects, one can use a cohort-sequential design

Question 14

Attrition effects

Answer 14

participants will drop out during the experiment. Some participants may be more likely to drop out, creating a confound.

Question 15

Practice effects

Answer 15

participants can improve on your measure for reasons other than age

Question 16

Instrumentation effects

Answer 16

the way you take your measurements may change over the years

Question 17

Cohort effects

Answer 17

because all your participants were recruited at the same time, they make up one cohort, with potentially unique aging effects