Brainscape's Knowledge GenomeTM

Browse over 1 million classes created by top students, professors, publishers, and experts.


See full index

PSYC4000 Advanced Methodology > Week 2 - Effect Sizes and Power Calculations > Flashcards

Week 2 - Effect Sizes and Power Calculations Flashcards

1
Q

Describe a correct decision (1-alpha), in that H0 is true within the population and we don’t reject H0

A

Finding no effect when there really was no effect in the world.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
2
Q

Describe a Type II error (beta), in that H0 is false within population but we don’t reject H0

A

Concluding there was no effect when there really was.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
3
Q

Describe a correct decision (1-beta) where H0 is false and we reject H0

A

Finding an effect when there really was one.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
4
Q

Describe a Type I error (alpha) where H0 is true and we reject H0

A

Concluding there was an effect when there really wasn’t one.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
5
Q

Pregnancy example of Type 1 error

A

Saying “You’re pregnant!” to a man

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
6
Q

Pregnancy example of Type 2 error

A

Saying “You’re not pregnant” to a pregnant woman

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
7
Q

Equation for power

A

1 - beta (i.e., type 2 error).

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
8
Q

Sample size definition

A

Number of participants required for a study

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
9
Q

Power definition

A

Ability for a study to find a true effect, if one is present within the population.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
10
Q

Alpha definition

A

probability of getting the observed effect due to chance (probability that we are willing to accept of making a type 1 error).

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
11
Q

Effect size definition

A

strength or magnitude of the effect between variables (relative to observation noise).

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
12
Q

4 types of design planning components

A
  1. Sample size
  2. Power
  3. Effect size
  4. Alpha
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
13
Q

Name for sample size calculation

A

a priori power calculation

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
14
Q

Name for power calculation

A

post hoc power calculation

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
15
Q

Name for alpha calculation

A

criterion analysis

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
16
Q

Name for effect size calculation

A

sensitivity analysis

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
17
Q

Difference between a one and two-tailed test (3 points)

A
  • Two tailed tests assess multiple scenarios (i.e., effects in different directions), one tailed assesses only one.
  • Two-tailed only half as strong because significance level split in half (2.5% each end) compared to one-tailed (5% one end) i.e., one-tailed has more power as assessing only one end.
  • Due to two-tailed only being half as strong, requires more participants to reach significance compared to one-tailed.
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
18
Q

Is a one-tailed or two-tailed test most common?

A

Two-tailed

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
19
Q

When might you use a one-tailed test

A

When there is strong evidence to suggest effect is directional.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
20
Q

Why do sample size calculators sometimes ask for summary statistics (means, SD’s etc)?

A

Because they can provide information on the distributions of the groups, and therefore effect size that needs to be detected. Sample size will be determined based on the distribution/effect size.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
21
Q

How would a sample size change if one group mean moved further away from the other groups mean? Why?

A
  • Sample size decreases
  • Moving away from other mean gives bigger effect size, therefore it is easier to detect the effect, so need less participants
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
22
Q

How would sample size change if both group means moved closer together? Why?

A
  • Sample size increases
  • Smaller effect size; harder to detect the effect, so need more participants.
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
23
Q

How does required sample size change as SD increases? Why?

A
  • Sample size increases
  • There is greater overlap or noise in the data, making it harder to detect an effect if there is one
24
Q

How does required sample size change as SD decreases? Why?

A
  • Sample size decreases
  • Less overlap, making the effect more apparent if there is one (less noise). Making it easier to detect the effect if it is there.
25
Q

How does changing number of measurements and number of groups affect sample size? Why?

A
  • Increasing number of measurements decreases sample size
  • Increasing number of groups increases sample size
  • This occurs due to variability
26
Q

Why is there a difference in required sample size to detect a main effect compared to detecting an interaction effect?

A

Main effect examines two groups (2x2 ANOVA), whereas interaction examines all 4 groups.
- Introduces more noise from doing more comparisons, therefore harder to find the effect, so need more participants to account for this

27
Q

All studies require a hypothesis after identifying the problem/research question - True or false?

A

False - not all studies require a hypothesis (sometimes they are exploratory).

28
Q

Does each type of research question have a superior design?

A

Partially true
- Some RQ’s can be addressed by multiple designs (depends on requirements/limitations of study to further refine)
- Other RQ’s clearly have a superior design.

29
Q

Study where changes or symptoms of a single person described

A

Case study

30
Q

Study where changes or symptoms of multiple people are described

A

Case series

31
Q

When might a case study/case series be used?

A

Might be used in clinical psychology or neuropsychology - it is common in medical contexts (rare conditions etc).

32
Q

Qualitative study

A

Analysis of non-numerical data obtained through interviews, focus groups, observation

33
Q

Aim of Qualitative studies

A

Identify themes related to a person(s) experiences

34
Q

Type of study that is descriptive (rather than drawing inference to large group of people) and is rich for hypothesis generation.

A

Qualitative study

35
Q

Type of studies that are observational (there are 3)

A
  1. Case-control
  2. Cross sectional
  3. cohort (longitudinal) studies
36
Q

Key criteria for observational study

A

No random assignment to groups - researcher cannot control the IV’s

37
Q

What does observational study do?

A

Compare an outcome(s) between naturally occurring groups - permits inference about relationships, but not cause and effect

38
Q

Common methodology for observational studies

A

Surveys, questionnaires.

39
Q

Type of experimental studies (2)

A

Controlled and field experiments

40
Q

What makes an experimental study?

A

experimenter intervenes on something e.g., random allocation to groups

41
Q

Type of study: participants randomly allocated to groups –> all participants in a condition receive same treatment/manipulation –> all participants have the same measurements taken.
Typically involves a control or baseline condition with no manipulation
Permits inference about cause and effect

A

Experimental study

42
Q

Quasi-experimental study

A

Where one variable is manipulated and another variable can’t be manipulated e.g., gender

43
Q

3 types of studies that come under a review study

A
  1. Literature review
  2. Systematic review
  3. Meta-analyses
44
Q

Review study that is a selective review

A

Literature review

45
Q

Why is there potential for bias in literature reviews

A

Because researcher selects articles to include

46
Q

Review study that aims to canvas all knowledge on a topic using replicable scientific methods

A

Systematic review

47
Q

Review study that takes key terms and searches databases, includes all papers and then narrows down using pre-determined criteria?

A

Systematic review

48
Q

Review study that is statistical analysis of similar outcome measures from previous studies and summarises the effect sizes in the literature

A

Meta-analyses

49
Q

6 factors that may compromise a research design

A
  1. money
  2. time
  3. access to populations
  4. access to different materials e.g., testing materials
  5. patient records access, surveys and computational records
  6. Access to knowledge
50
Q

What is meant by “compromise does not mean compromised”

A

Identify whether limitations prevent the research design from addressing the hypothesis/research question
Plan for your research design to maximise your chances of detecting an effect if there is one (power calculation).

51
Q

What are all of these definitions referring to?
- Probability of correctly rejecting the null hypothesis
- Probability of rejecting the null hypothesis when the null hypothesis is false
- Probability that a statistical test will detect an effect that is present
- Probability of avoiding a Type 2 error
- Chances of detecting an effect if there is one

A

Statistical power

52
Q

Review page 5 of week 2 lec notes

A

Review page 5 of Week 2 lec notes

53
Q

Effect sizes are notoriously difficult to estimate. Why are effect sizes from published studies not always the best option?

A

Published studies are notoriously unreliable - need to find different sources for effect size estimate because studies done for first time can overestimate the effect size.

54
Q

5 ways to find an effect size

A
  1. Pilot data
  2. Previous study similar to yours
  3. Meta-analysis (good source if available as gives aggregate effect size across numerous studies and removes noise that is associated with calculating individual studies’ effect sizes
  4. Convention (cohen’s d)
  5. Other e.g., real life –> do you think effect would be small based on what seen in real life (where derived RQ from?)
55
Q

Purpose of a priori power calculation

A

To determine the minimum sample size required for a particular statistical test.

56
Q

Which has higher power - parametric or non-parametric?

A

Parametric - assumes residuals fall within a normal distribution

PSYC4000 Advanced Methodology (2 decks)

Brainscape helps you reach your goals faster, through stronger study habits.
© 2024 Bold Learning Solutions. Terms and Conditions