2. Personality Test Construction
Goal:
Gain an increased understanding of the
concepts reliability and validity as they
pertain to tests
Gain an increased understanding of test
development methods
3. Test Construction Procedure
1.
2.
3.
4.
5.
Identify a need for a new test
Assemble an item pool (decide on
scale and item formats)
Pilot item pool
Select “good” items
Examine test’s psychometric properties
(reliability and validity)
4. 1. Identify Need for a New Test
What
is the objective of the new test/is
there really a need for it
How will the test be administered?
What is the ideal item format for this
test?
Should more than one form be
developed?
What special training will be required of
test users in terms of administering or
8. 3. Pilot Item Pool
Try
the pool of items out on people for
whom the test is being developed
Test should be administered under
conditions similar to those that the
developed test will be administered (e.g.
same instructions, time frame, time
limits)
9. 4. Select “Good” Items
Selecting “good” items involves complex
statistical analysis of the test results
which varies according to the purpose of
the test.(called item analysis)
However, in tests of attitudes or personality
characteristics one consideration is
whether individuals endorse the full
range of the scale provided.
10. 5. Examine Test’s Psychometric
Properties
Does
the test yield consistent results
(reliability)?
Do the test items measure the intended
construct (validity)?
12. Test Construction Exercise:
Procedure
Divide into groups of 4 to 5 students
In Class
As a group, develop an item to distinguish first
born from later born children
Note: use a personality construct and not a
physical characteristic (e.g. I have no older
siblings)
Develop two responses for the item
Once your item is ready, tell Sara or Eunyoe
so they can write it on the board (so others
won’t give the same item)
16. Reliability
Consistency
of the observations or
measurements
Reliability is inversely related to the
degree of error in the instrument.
High measurement error translates to
low reliability
Low measurement error translates to
high reliability
17. What !?
What does this mean!?
High measurement error
translates to low
reliability
Low measurement error
translates to high
reliability
Easy Example: A broken
scale
There will be high
measurement error on a
broken scale, correct?
How consistent are the
weights likely to be on a
broken scale?
Is a broken or working
scale going to have
more error?
Is the broken or working
scale going to be more
reliable?
18. Types of Measurement Error
Random
Factors unpredictably
influence
measurements.
Systematic
A persistent bias in the test
or in the interpretations
made by examiner.
Examples:
Mood, environmental
distractions, hunger or
motivation interfere with
the responses.
Systematic errors, because
they are consistently
made will not affect
reliability but they will
affect validity
19. Types of Reliability
Inter-rater
reliability (relevant to
observational systems and psychological
assessments requiring ratings or
judgment)
Test-retest reliability
Split-half
Note: Each form of reliability is not equally
important for every assessment method
20. Inter-rater Reliability
Degree of correspondence between two raters
Inter-rater reliability of diagnoses based on
DSM criteria improved with DSM-III and the
development of operational criteria for most of
the mental disorders
Note: We will learn how to calculate next week!.
21. Test-Retest Reliability
The
consistency of results over periods
of time.
The
consistency of the results for a test
given at two different time periods
The
correlation of test result scores
22. Quantifying Test-Retest Reliability
Reliability is expressed as a correlation
coefficient
Values range from 0 (not at all consistent or
reliable) to 1 ( perfectly consistent and reliable.
The value for adequate reliability is about .80
or greater
23. Factors Affecting Test-Retest
Reliability Estimates
Length of the intervening interval
Stability of the measured trait
For example:
In characteristics that are stable, like intelligence, the
interval of time between the two tests should not affect
the stability of the results.
In contrast, in characteristics that are not stable, like
depressed mood, the longer the interval between tests,
the less reliable or consistent the scores. (not necessarily
bad)
25. Validity
A test can be reliable (consistently give the
same results) but not valuable.
Why?
If the test does not measure the correct
construct, then it is not useful even if the
results are consistent.
27. Types of Validity
Face
validity
Content validity
Criterion validity (predictive and
concurrent)
Discriminant
Construct validity
28. Face Validity
A judgment about the relevance of test items
A type of validity that is more from the
perspective of the test taker as opposed to the
test user
Example: Personality tests
Introversion-Extroversion test will be perceived
as a highly (face) valid measure of personality
functioning
The inkblot test may not be perceived as a (face)
valid method of personality functioning
29. Content Validity
Degree
to which the measure covers the
full range of the (personality) construct.
and
Degree to which the measure excludes
factors that are not representative of the
construct
30. Criterion Validity
The
degree to which the test results
(from your measure) are correlated with
another related construct.
WHAT!?
For example: the degree to which scores
on an intelligence test are correlated with
school performance or achievement.
31. Types of Criterion Validity
Concurrent: the two constructs are assessed at the same
time
Predictive: one construct may be measured at a later
date
For example:
Concurrent: the correlation of SAT score with G.P.A. at the
time of taking the SAT in high school.
Predictive: the correlation of SAT score taken in high school
with final G.P.A. upon graduating from college
32. Discriminant Validity
The degree to which the score on a measure
of a personality trait does not correlate with
scores on measures of traits that are unrelated
with the trait under investigation.
For example: (from text)
Trait being measured: phobia
Unrelated trait: intelligence
You would not expect the score on your phobia
scale to be correlated with the score on an
intelligence test
33. Construct Validity
The
degree to which the measure
reflects the structure and features of the
hypothetical construct that is being
measured
Measured by combining all these
aspects of validity.
34. Exercise: Reliability and Validity applied to the
Edinburgh Postnatal Depression Scale (EPDS)
Let’s
consider reliability and validity in
the context of a real measure: the EPDS
35. What is the Edinburgh Postnatal
Depression Scale (EPDS)?
John Cox, Jenifer
Holden & Ruth
Sagovsky
10 item depression
screening tool
(reliable and valid)
Simple to complete
Acceptable to
mothers and health
workers
36. What is the Edinburgh Postnatal
Depression Scale (EPDS)?
Psychometric Characteristics
10 item scale
Assesses mood aspects of depression
not confounding somatic symptoms
Acceptable to women
Validated
Translated into many languages
37. Stems of all 10 EPDS Items
I have been able to laugh and see the funny side
of things.
I have looked forward with enjoyment to things.
I have blamed myself unnecessarily when things
went wrong.
I have been anxious or worried for no good reason.
Things have been getting on top of me.
38. Stems of all 10 EPDS Items
(cont)
I have felt scared or panicky for no very good
reason.
I have been so unhappy that I have had
difficulty sleeping.
I have felt sad or miserable.
I have been so unhappy that I have been
crying.
The thought of harming myself has occurred to
me.
39. Psychometric Evaluation of the
EPDS: An Exercise
Is
the EPDS a good measure of
depression?
Psychometrically, what does it mean to ask
if the EPDS is a “good” measure of
depression?
Note: Follow the questions on the handout
41. Test Construction Exercise:
Part 2: Evaluating Developed Tests
Regroup into your “test groups”
2. Evaluate items in terms of content
validity and adequacy of scales
3. Select final items for test
4. Propose methods for evaluating
reliability and validity of new measure
1.