STATS 10-Research designs

Question 1

Quasi-independent experimental variables

Answer 1

• Characteristics that can’t be randomly assigned
• Often properties already existing in subjects
  
  • e.g. gender, handedness, age group, hair colour, symptoms
  • Best you can do is split into groups

Question 2

True experimental variables

Answer 2

• You can control these in a true experiement
• You can randomly assign people to groups
  
  • Drug treatmnet versus counselling
  • Meal A,B,C,D
  • Drug A or B
  • Using calculator or using a slide rule
    *

Question 3

Independent and dependent variables… think

Answer 3

• Which do we try and control in an experiment
• Which do we want to measure
• See the podcast: we’ll play spot the variables

Question 4

Experimental methods 1

Answer 4

• A method of collecting data which allows the researcher to make causal inferences about the relationship between 2 or more variables

Question 5

Experimental methods 2

Answer 5

• The researcher manipulates one or more independent variables to see the effect on the dependent variables
• e.g. Effect of alcohol on memory
  
  • IV: Amount of alcohol consumed
  • DV: Score on a memory test

Question 6

Types of experimental design 1

Answer 6

• True experiemental design: Also known as randomised design researchers can randomly assign participants to different experiemental condition
  
  • E.g. Assign “Normal” participants to groups that consume different amounts of alcohol

Question 7

Types of experimental design 2

Answer 7

• Quasi-experimental design
• Similar to experimental design but the researcher is unable to randomly assign participants to groups
  
  • e.g. Compare pre-existing alcohol consumption groups: Heavy vs Light drinkers or alcoholics vs non-alcoholics

Question 8

Confounding variables

Answer 8

• If the groups to be compared differ in ways other than which the researcher has manipulated, the experiment has confounding variables
• e.g. Weight changes how fast alcohol affects people
  
  • Can lead to incorrect conclusions

Question 9

Variables so far

Answer 9

• Independent variables (Condition)
• Dependent variables (Measured)
• Confounding variables- what gets in the way
• How can we control confounding variable

Question 10

Randomisation- why randomise

Answer 10

• Ensures that each participant is equally likely to be assigned to a given condition
• WHY RANDOMISE
  
  • Prevents experimenters (un)intentionally biasing their results
  • Distributes the occurrence of potential moderating/confounding variables equally among the experimental conditions
  • Enables the use of powerful statistical tests the can help determine the causal relationship between variables

Question 11

Two ways to compare groups/Conditions

Answer 11

• We use the term “condition” for the independent variable
• Independent groups (between-subjects) design
• Repeated measures (within-subjects) Design
• Mixed design

Question 12

Independent group design- between subjects

Answer 12

• Examples: effects of alcohol consumption (IV) on short term memory performance (DV)
• Randomly assign participants to one of the 2 groups
• Administer alcohol accordingly
• Measure their memory performance and compare

Question 13

Independent group design- potential problems

Answer 13

• How do we ensure that any difference in memory function result from alcohol intake rather than some other factor
  
  • E.g. Age, gender, nutritional factors, confounding variables
• How do we account for the possibility that some of our participants have more experience with memory tests
• We CAN’T, but we can minimise these effects by randomly assigning participants to experiemental conditions (or levels of the IV)

Question 14

Matched groups design

Answer 14

• An answers to make sure that subjects in both groups are matched as closely as possible on potential confounding variables
  
  • Age, gender, Experience with alcohol

Question 15

Repeated Measures Design- within subjects

Answer 15

• Example: effects on alcohol consumption (IV) on short-term memory performance (DV)
• Participants now take part in both conditions (levels) of IV
  
  • Test before alcohol
  • Administer alcohol accordingly
  • Test after alcohol

Question 16

Repeated measure Design: Potential problems

Answer 16

• Requires fewer participants than independent groups design
• Each participant served as their own control
  
  • Removes some confounding factors
• BUT
  
  • Order effects
  • Once participants have been exposed to one level of the IV there no way to return them to their original state
    
    • Practice effects, fatigue effects

Question 17

Counterbalancing: An answer

Answer 17

• Problem- in repeated measures design, the act of measuring the DV introduces order effects
• Solution- counterbalancing (randomly assigning the order)

Question 18

Example: effects of alcohol consumption (IV) on short-term memory performance (DV)

Answer 18

• Randomly assign participants into the group a or Group B
• Alternate the sequence of administration of experimental conditions (ABBA)
• Order effects eliminated but should be common to both condition
  *

Question 19

ID the experimental design- IV; IV levels; DV; Design; True/Quasi

Answer 19

• An investigator is interested in whether males or females are better at the computer game Tetris
• He measures the scores of 30 males and 30 females playing tTetris
• What is the IV
• What is the DV
• What is the subjects design of the study
• Is it a true or quasi-experimental design

Question 20

ID the experimental design

Answer 20

• An investigator is intrestedin whether easting fish improves exam scores
• She instructs 1/3 of her students to eat fish
• 1/3 to eat red meat
• 1/3 to eat no meat the night before the exam
• What is the IV
• What is the DV
• What is the subject design of the study
• Is it a true or quasi-experimental design

Question 21

Causation

Answer 21

• Internal validity- How correct is it to claim that treatment X causes outcome O
• One-group case study (X-O)
• One-group pre-post design (O-X-O)
• No comparison groups or non-treatment control group. How can you tell?

Question 22

Pre-experimental design

Answer 22

• One-group post-test (X-O) design: one group of participants is exposed to one level of an IV and then observed (or measured)
  
  • e.g. measure the effect of alcohol on memory for one group at only one occasion
• One group pre-post-test design (O-X-O) one group of participants is given a pre-test then a treatment and then a post-test
  
  • Measure the effects of alcohol on memory by comparing pre-test and post-test performance

Question 23

Spot the experimental design

Answer 23

• An investigator is interested in whether an audience’s rating of a bands performance is greater when the performance is being recorded for live broadcast compared to when it is not
• She is also interested in whether the type of music has an effect
• The investigator asks a group of 50 audience members at an arctic monkeys concert to rate their enjoyment of a televised song and a non-televised song
• At a seperate concert she asks a group of 50 girls aloud audience members to rate their enjoyment of a televised song and non-televised song

Question 24

ID the experiment in the previous slide

Answer 24

• What is the experimental design
• What is the first IV
  
  • How many levels
  • What is the subject design
• What is the second IV
  
  • How many levels
  • What is the subject design
• What is the DV
• How would you design a better experiment

Question 25

ID the experimental design

Answer 25

• An investigator is intrestedin whether students reaction times on driving task are slower when they are hung over than when they have not consumed alcohol the previous night
• She measures the reaction times of 50 students the morning after drinking and the same students reaction times the morning after an alcohol free video night

Question 26

From the previous slide

Answer 26

• What is the IV
  
  • Levels
• What is the DV
• What is the subject design of the study
• What is the confound
• How would you change the research design to remove the confound