Wk2: Constructs

Question 1

Three validities for causation

Answer 1

• Construct
• Internal
• External

Question 2

Define construct validity

Answer 2

Whether the construct is operationalised through the msmt
How much the concept is reflected in the msmt used
Variable true to itself
E.g. Wealth~Hrs worked = low; Wealth~ #assets and annual pay = high

Question 3

Define reliability

Answer 3

The replicability of the msmt tool used
The accuracy and precision of the msmt tool used
E.g. measuring height using 1m stick (low) or 1m ruler (high)

Question 4

Types of reliability

Answer 4

• Test-retest
• Split-half
• Parity

Question 5

Define test-retest reliability

Answer 5

Correlation b/w scores at two time points
When a test taken multiple times produce the same results in individuals
Cons: Susceptible to learning
Cons: Works for static, not dynamic constructs

Question 6

Define split-half reliability

Answer 6

Correlation b/w scores in both halves of experiment
When individuals tests’ are halved and the correlation between the two halves are measured.
Cons: Overtesting, test length

Question 7

Define parity reliability

Answer 7

Correlation b/w scores in odd and even trials
When individuals tests’ are seperated into odd and even trials and the correlation between the two halves are measured.

Question 8

Define criterion validity

Answer 8

How well the test reflects the criterion it aims to test
Empirical validity
E.g. How well the simulated surgery scores reflect doctors’ actual surgery outcomes. Testing for actual surgery outcomes.

Question 9

Define content validity

Answer 9

How well the test appears to reflect the construct
Opinion validity
E.g. How much of the lecture content is present during the exam. May test knowledge of course through tutorial content only, but not wholly content-valid.

Question 10

Define convergent validity

Answer 10

How well other correlates of the construct correlate with the test
Do the scores of related correlates converge?
E.g. How much Psyc3020 and stats grades correlate with Psyc3042 grades

Question 11

Define divergent validity

Answer 11

How well other tests of the construct do NOT correlate with the test
Do the scores of non-related concepts diverge in magnitude?
E.g. How much dep scores correlate with anxiety - shouldnt correlate
How much hrs studying reflects alcohol consumption - should not correlate

Question 12

Define common method variance

Answer 12

It is common for the methods used to show a variance in results when the construct should not.

The variance we see in test results that shouldn’t be present in the construct we try to measure
E.g. self-reported alcohol inebriation (“how drunk do you feel”) vs. # pints drunk vs. BAC vs. metabolism (BAC over time). Each test reflects different systematic variances of the method, not the results.

Question 13

What are multi-trait, multi-method techniques?

Answer 13

Measuring convergent and divergent validity over multiple methods (self-report, observation, behavioural msmts, etc.)
Usually 2 constructs with 2 methods,
E.g. Assertiveness and aggression through self-report and observation

Question 14

Define internal validity

Answer 14

Whether the msmt reflects a causal relationship between constructs
Variable true to its relationship

Question 15

Failures of internal validity

Answer 15

• Spurious scoring
• Scoring and data analysis errors
• Direction error

Question 16

Define spurious scoring

Answer 16

Coincidence
E.g. US Industrial manufacturing doctorates awards ~ US pc mozzarella cheese consumption, r=0.9

Question 17

Define scoring and data analysis errors (internal validity)

Answer 17

The analysis misconstrues the results; misinterpretation.
E.g. A vaccine decreases fatalities in a disease from 23 in 1000 to 9 in 1000. Is this a 60% decrease or a decrease of 0.014%?

Question 18

Define reverse causation (internal validity)

Answer 18

Assume X causes A when A causes X
E.g. Assume married people are happier because they married; actually happier people are the ones who marry.

Question 19

Determining causation through experimentation

Answer 19

• Manipulate IV (true experimental design)
• Determine preceding variable (A precedes B)
• Determine mechanism of causation (why A causes B)

Question 20

Define mediation

Answer 20

A -> C because A -> B -> C

Question 21

Mediation vs mechanism of causation

Answer 21

Mediation is through another independent construct, whereas the mechanism doesn’t
E.g. mediation: guns cause aggression through feeling threatened
E.g. mechanism: studying improves grades because studying is the process of improving understanding and learning, and thus grades.

Question 22

Measuring mediation

Answer 22

• HMR: model 1 (X->Y) to 2 (X->Y, cont.M; M->Y, cont.X)
• Mediation: X->Y, cont.M weaker than X->Y

Question 23

Full mediation

Answer 23

• Where X->Y cont. M (2,1) is non-significant
• The path X->Y determined by M

Question 24

Partial mediation

Answer 24

• Where X->Y cont. M (2,1) is significant, but weaker than X->Y (1)

Question 25

Define external validity

Answer 25

Whether the msmt and sample reflects the other populations, settings, and times. AKA Generalisability Variable true to the population

Question 26

Failures of external validity

Answer 26

* Testing on a subpopulation, applying findings to whole population. E.g. testing meds on males, assume safety, causes birth defects * Results not replicable with alternative stimuli

Question 27

Define moderation

Answer 27

A->C proportional to B E.g. greater B = stronger A->C correlation

Question 28

Moderation vs external validity failures

Answer 28

External validity failures are misattributions of findings, moderation is the correctly identified attribution of the change in the findings (∆r≈B)

Question 29

Moderation vs mediation

Answer 29

* Moderation not correlated with X or Y * (∆r≈B) * E.g. Hrs in gym (X) ≈ Gym satisfaction (Y), *moderated* by nutrition (B) * Mediation correlated with X and Y * (X≈B≈Y) * E.g. Hrs in gym (X) ≈ Gym social interaction (B) ≈ Gym satisfaction (Y)