Lecture 4 ID&E Assessment and Evaluation
25-09-19
Program
Why might this be necessary. Let’s look back on history
Predictive validity: if you can make an IKEA chair, you can work there.
- There was a need to change validity
Psychometrics
- We try to measure what is in someone’s head
- Assumption that I already know the psychometrics terms
- Using data to get an estimate of how well the mastery of the behavior is & to make an
assessment tool more reliable.
- Make explicit choices when designing assessments
Edumetrics
- What consequences are from your tool and score?
- Passing or failing?
- Higher stake higher validity and reliability necessary?
,Nature assessments (performance assessment) do we need psychometrics because a natural
assessment is not inside someone’s head like latent variables.
Criterion-based assessment using rubrics to actually assess someone’s behaviour.
Factor analysis do not apply to knowledge tests
No matter how good you design your data, there will always be some kind of error and items which
would lower the overall score. And how do you deal with it?
If you look at that formal arguments (data), should we look at more than that (make your choices
more explicit)? Construct validity is the most important one.
Messick: before data arguments are useful, we need more than that. He designed the validity
measures, but later stated that we need more than that.
, Interpretive validity argumentation
Lawyer. Example drivers license
Data: Poor parking, smoked a cigarette, speed limit violation
Claim: Failed license exam
Warrant: (interpretive arguments). The warrant is actually supporting the relationship between the
data and the claim. Certain arguments of the decision of the claim based on the data concerning:
relevance of the data (explicate the claims), sufficient time, reliable scoring, proper motor cycle.
Backing: (validity arguments) the evidence for the warrant: Certification board (organized meeting in
which experts decide what should be measured and what counts as passing/failing. Nation-wide
criteria?), time measured, trained examiner (reliable scoring/coding, video-taping and showing to a
second examiner), certified motor cycle. Defied in: procedural evidence, internal evidence and
external evidence.
Reservation/rebuttal: emergency situation (you had to speed to give space for the police),
exceptions, when the line of reasoning does not apply.
Qualifier: how we qualify this way of reasoning (data claim, evidenced by warrants, supported by
backing, without rebuttals). More warrants with more backing probably gives a higher quality.
25-09-19
Program
Why might this be necessary. Let’s look back on history
Predictive validity: if you can make an IKEA chair, you can work there.
- There was a need to change validity
Psychometrics
- We try to measure what is in someone’s head
- Assumption that I already know the psychometrics terms
- Using data to get an estimate of how well the mastery of the behavior is & to make an
assessment tool more reliable.
- Make explicit choices when designing assessments
Edumetrics
- What consequences are from your tool and score?
- Passing or failing?
- Higher stake higher validity and reliability necessary?
,Nature assessments (performance assessment) do we need psychometrics because a natural
assessment is not inside someone’s head like latent variables.
Criterion-based assessment using rubrics to actually assess someone’s behaviour.
Factor analysis do not apply to knowledge tests
No matter how good you design your data, there will always be some kind of error and items which
would lower the overall score. And how do you deal with it?
If you look at that formal arguments (data), should we look at more than that (make your choices
more explicit)? Construct validity is the most important one.
Messick: before data arguments are useful, we need more than that. He designed the validity
measures, but later stated that we need more than that.
, Interpretive validity argumentation
Lawyer. Example drivers license
Data: Poor parking, smoked a cigarette, speed limit violation
Claim: Failed license exam
Warrant: (interpretive arguments). The warrant is actually supporting the relationship between the
data and the claim. Certain arguments of the decision of the claim based on the data concerning:
relevance of the data (explicate the claims), sufficient time, reliable scoring, proper motor cycle.
Backing: (validity arguments) the evidence for the warrant: Certification board (organized meeting in
which experts decide what should be measured and what counts as passing/failing. Nation-wide
criteria?), time measured, trained examiner (reliable scoring/coding, video-taping and showing to a
second examiner), certified motor cycle. Defied in: procedural evidence, internal evidence and
external evidence.
Reservation/rebuttal: emergency situation (you had to speed to give space for the police),
exceptions, when the line of reasoning does not apply.
Qualifier: how we qualify this way of reasoning (data claim, evidenced by warrants, supported by
backing, without rebuttals). More warrants with more backing probably gives a higher quality.
