-
Is operational definition testworthy
- ex. you can use a measurement of your cake-baking skill as an operational definition of you intelligence. clearly this is not so.
- This definition lacks reliability and validity
-
Face validity
whether something looks valid on the surface
-
the three sta tistical elements of testing
- Really Very Nasty
- R= Reliability
- V= Validity
- N= Norms
-
Norms
gives you a reference point
-
Validity
establishes that our measurement measures what its supposed to measure
-
Reliability
- establishes that our measurement is stable or consistent (reliable)
- the correlation betwee two measurements of the same thing
- 1 means perfect reliability
- we want them to be high .6-.8
-
Operational definition
turns our abstract contruct into something we can measure
-
abstract construct
the thing that we are really interested in but we cant measure it directly
-
why are longer test more reliable
- each person answers questions in a variety of ways. if you only ask one quesrion, you might not get a true picture of that person.
- each item test what you want it to and plus some other nonsense
- when you average them the nonsense gets averaged out leaving only the construct of interest.
-
two kinds of reliability
- intrarater reliability: test the same person more than once
- interrater reliability: test different people
-
Intrarater reliability
- retest the person on different occasions
- problems: memory effects and practice effects
- internal consistency: this is coceptually different, because we are not retesting in any way
-
Internal consistency
- unidimensionality (single dimension)
- only makes sense when all of the items are supposed to be measuring the same thing.
-
split-half reliability
- aplit the test in hald, and correlate one half with the other half
- this is like treating one test as two
- it divides the test in half, which reduces the reliability
- problems: you could get many different scores, depending on how you split it
-
Cronbach's Alpha
- correlate each item on the test with the total test score
- average them
- there is only one
- Alpha is the sign you will always see in reliabilty
-
Kuder-Richardson
like alpha but only applies to a test where there are right and wrong answers
-
spearman-brown reliability
an adjustment formula, to increase the reliability back to what it would be for the full-length test
-
Interrater reliability
- only applies when we expect people to agree with each other
- correlate the judges with each other
- ex. raters of videotapes
- doesnt apply to paper and pencil (objective) tests, but does apply where tests have subjective elements, such as projective tests
|
|
