Permalink
Switch branches/tags
Nothing to show
Find file
Fetching contributors…
Cannot retrieve contributors at this time
57 lines (55 sloc) 2.85 KB
Important Aspects of Validity
* Validity is inferred, not measured
With reliability, you measure. However, with validity you don't
* Validity depends on many different types of evidence
* Content
* Criteria
* Construct
* Validity is expressed by degree
Nothing is complete invalid or valid
* Validity is specific to a particular use
Only really measures a single construct
* Validity refers to inferences drawn, not the measure itself
* Validity is a unitary construct
For a long time validity was split into 3 (content, criteria, construct). Now it's seen as a single thing with 3 parts
Types of Categories of Validity
* Content-related Validity
* Adequate Measurement
Has to get at what we want it to get at
* Relevant
* Representative
* Face Validity
* Assesses whether the measurement items appear to measure what they are suppposed to measure.
e.g., Really badly worded questions.
* Do the assessment items look valid?
* Criterion-Related Validity
DV-related validity. See if our measure of whatever construct relates to other measures of other constructs related to ours
e.g., SAT test and GPA. Does the SAT predict GPA of freshman year. If they're related, then we have some Criterion-related validity
* Types of Testing
* Concurrent
Take both tests, and then see how they both correlate with GPA
* Predictive
Have a theory that Fear of Failure/Achievement Motivation leads to Executive Success. Measurement would be average time between promotions. So with a survey used to determine FoF/Ac motivation you predict average time between promotions
* Types of Criterion
* Construct Validity
The degree to which your manifest variable measures your hypothetical construct
How to get to it
* Formulating hypotheses about the links between the operational measure and the hypothetical construct
* Conduct a study to determine the empirical relationship between your measure and other measures of other related constructs
* Interpret the results in terms of construct-related validity
* Kinds of Construct Validity
* Convergent Validity
* Discriminant Validity
* Method Variables
* Substantive Variables
* Manifest Variable
* Differential Validity
* When a measure is more valid for one group than the other.
* The relationship between Reliability and Validity
* For a measure to be valid, it has to be reliable
* However, reliability is not sufficient to claim validity.
A scale. If the scale is always 2 lb heavier, then it's reliable, however it's not valid.
Threats to Internal Validity
* The extent to which one can be confident that the results are reported support the casual hypothesis being tested
* It is the basic minimum without which any experiment is uninterpretable
* Are variables that appear to be causally related, REALLY causally related, or is this relationship spurious