Chapter 5-construct validity-MTMM Flashcards
MTMM def
Multitrait-Multimethod Matrix.
Way of representing the relations between several constructs and several methods for measuring these constructs.
Why do we Need a MTMM? (3)
(1) Both convergent and divergent evidence of validity is needed.
(2) Each test is a construct-method unit: a combination of construct content and measurement procedures not specific to that content
(3) To examine divergent validity, convergent validity, and method variance, MORE THAN 1 CONSTRUCT & MORE THAN 1 MEASUREMENT METHOD (minimum 2) are needed in the validation process
What does Method variance represent?
SYSTEMATIC error
Characteristics of method that will influence how responders will respond to questions important for our attributes
Test score variance is composed of ____ (3)
True score variance + Method variance + Random error
MTMM presents all the interrelations resulting when__________
each of several constructs is measured by each of several methods
If everything is good, we’re looking for: (3)
(1) High “trait variance”
(2) Low “method variance”
(3) Low “irrelevant variance”
Trait variance def
Variance due to construct
Method variance def
Variance due to method
Irrelevant variance def
Variance shared with theoretically unrelated measures
Minimum requirement of MTMM
At least 2 measures representing 2 different methods for each of 2 constructs.
MTMM: Regions (6)
(1) Monomethod block
(2) Monotrait-monomethod values
(3) Heterotrait-monomethod triangle
(4) Heteromethod block
(5) Monotrait-heteromethod values
(6) 2 Heterotrait-heteromethod triangles
What are the 2 diagonals that exist?
Reliability and Validity diagonals
What is the Reliability diagonal?
Monotrait-monomethod values (in monomethod block).
Tell how reliably each construct (A, B, C) can be measured with each method.
What is the Validity diagonal?
Monotrait-heteromethod values
Tell how well a construct is measured using different methods.
-> CONVERGENT validity coefficients
Rules for Evaluation of an MTMM (4) - expectations
(1) Validity diagonals (=convergent validity) should be HIGH
(2) Heterotrait monomethod values should be LOWER than convergent validity
(3) Heterotrait heteromethod triangles: LOW
(4) Reliability diagonal - monotrait-monomethod correlations: High/as expected