Validity Flashcards
1
Q
How would you define validity?
A
- The appropriateness, meaningfulness, and usefulness of the specific inference made from test scores
- i.e., the process of accumulating evidence to support such influences
2
Q
What’s another definition for validation?
A
- The scientific inquiry into test score meaning
3
Q
What’s content validity?
A
- Determine whether a test forms a representative sample of the defined content domain
- A clear definition of the content domain is necessary
- Ex. Midterm assessing students on every topic that was covered in lecture
4
Q
How can you determine whether a test has a representative sample of the content domain?
A
- Expert’s systematic judgement (i.e., inter-rater reliability)
- No statistical measure of content validity
5
Q
What’s face validity?
A
- The extent to which a appears to be reasonable to those who take or use it.
- Not a ‘real’ form of validity, but still important for the participant to perceive that there is high face validity
- Possible to have low content validity but high face validity
6
Q
What’s the Bandwidth-Fidelity dilemma?
A
- Bandwidth - the amount of info that is contained in a message
- Fidelity - the accuracy with which that info is conveyed
- When creating test items, often have to have a trade-off between these two constructs
- Higher fidelity = higher reliability at the cost of having a narrow range of assessments.
7
Q
What’s criterion-related validity?
A
- Involves the empirical relation between the measure and variables having a given level of relevance to the attribute
8
Q
What type of validity are correlation coefficients?
A
- They’re criterion-related validity
9
Q
What’s convergent validity?
A
- Indicates that a test should correlate strongly with theoretically relevant variables
10
Q
What are the two types of convergent validity?
A
- Predictive validity - data collected over two time periods
- Concurrent validity - Already have predictive scores, now assessing the outcomes of the test used. Often used more than predictive validity as it is less expensive/less time-consuming
11
Q
What are range-restrictions?
A
- When only looking at a specific range of data can actually lower your correlation coefficient when not taking into account the whole range of values (i.e., not looking at the entire slope)
- Seen in job application processes when only looking at the performance reviews of those hired.
12
Q
How can we correct for the changes in reliability caused by range restrictions?
A
- Use the correction for attenuation formula
- Must use to recover the full correlation
13
Q
What’s construct validity?
A
- Refers to the overall evidence that the test measures the variable that it is intended to measure, rather than anything else
14
Q
How can constructs be expressed/described?
A
- Can use nomological networks to help describe what the constructs are
15
Q
T/F: The construct validation process is a lifelong process, and perhaps one of the ultimate goals of psychological science
A
- TRUE
