8. Two primary WJ IV COG design decisions
Gc
Gf
Gwm
Oral
Vocabulary
Number
Series
Verbal
Attention
General
Information
Concept
Formation
Numbers
Reversed
AnalysisSynthesis
Memory for
Words
Glr
Gv
Story Recall
Visualization
Visual-Aud
Learning
Picture
Recognition
Ga
Phon.
Gs
Processing
Letter-Pat.
Match.
Nonword
Pair
Repetition
Cancellation
Number-Pat
Match.
Obj-Num.
Sequencing
• Which 7 tests should be combined for the GIA (g) cluster?
• Which 2 tests from each CHC factor domain should be
combined for the 7 CHC factor clusters?
13. Select concurrent validity evidence:
Correlations of WJ IV primary COG g-scores
with external measures
WISC-IV WAIS-IV WPPSI-III KABC-II
FS IQ
FS IQ
FS IQ
FCI
(n =174) (n =177) (n = 99)
(n=50)
SB-5
FS IQ
DAS-II
GCA
(n = 50)
(n = 49)
WJ IV g-measures
(see prior note regarding technical info presented at NASP that can’t be released
yet until tech. manual published)
Conclusion: The WJ IV GIA, BIA and Gf-Gc composite clusters
demonstrate strong validity evidence as measures of general
intelligence when the criterion are the global composite/total
scores from other major IQ batteries in the field
15. 1-factor (unrotated) common-factor solution for WJ IV
COG / WISC-IV composite scores (n=173)
Verbal Comprehension Index (VCI)
Auditory Processing (Ga)
Fluid Reasoning (Gf)
Perceptual Reasoning Index (PR)
Comprehension-Knowledge (Gc)
Short-Term Work. Memory (Gwm)
Working Memory Index (WMI)
Long-Term Retrieval (Glr)
Visual Processing (Gv)
Processing Speed Index (PSI)
Cog. Processing Speed (Gs)
Loadings on Communality
first unrotated
Estimates
common factor
0.809
0.654
0.804
0.646
0.804
0.646
0.800
0.639
0.779
0.607
0.764
0.584
0.749
0.562
0.683
0.466
0.604
0.365
0.569
0.323
0.537
0.288
16. •The WJ IV GIA score is as
good (better?) a measure of
general intelligence (g) as
the WISC-IV FS IQ when
defined by g-loadings and
MDS cognitive complexity
analysis.
Measures closer to the center of
the radex of more cognitively
complex
2
1
GWM
GV
WMI
PRI
FSIQ
PSI
0
• The WJ IV Ga cluster is a
measure of complex
cognitive abilities;
comparable to WJ IV &
WISC-IV Gf/PRI composites.
GF VCI
GIA
GA
GLR
-1
GC
GS
-2
-2
-1
0
1
2 MDS solution for WJ IV / WISC-IV
composite and g-scores (n=173)
2
•The WJ IV measures
cognitive abilities not
represented in the WISC-IV
(Ga, Glr, and possibly Gv).
20. Five primary design principles for WJ IV COG GIA
The WJ IV COG GIA cluster tests should:
1. Be the best factor indicators of each CHC broad domain
2. Be the best predictors of achievement from each CHC broad
domain
3. Be the most cognitively complex indicators from each CHC
broad domain
4. Be the best measures of g (general intelligence) from each
CHC broad domain
5. Collectively should have a relatively equal balance of type of
stimulus characteristics (verbal, numeric, figural)
Related
22. Cognitively complex design characteristics
(Lohman, 2011)
• Larger number of cognitive component processes
Parameters
of cognitive
efficiency
in info.
proc.
models
• Accumulation of speed component differences
• Increased demands of attentional control (AC) &
working memory
• More important component processes (e.g.,
inference)
• More demands on adaptive functions (assembly,
control, and monitoring).
28. Professor Butts and the “over
engineered” self-operating
napkin machine (aka., Rube
Goldberg machines)
29.
30. COG Test 5: Phonological
Processing
• Ga (PC) / Glr (LA/FW)
•3 subtests (Word Access; Word Fluency; Substitution
• Measures three aspects of speech sound processing
that requires the efficiency construction of soundbased lexical representations
• High in cognitive complexity and g. Best single Ga
test predictor of achievement. High loading on Ga
and secondary low loading on Gc (accessing the
lexicon). Also loaded on narrow LA factor in
broard+narrow bottom-up CFA models.
• In GIA, Ga, and all reading and writing scholastic
aptitude clusters
31. COG Test 5A: Phonological ProcessingWord Access
• Word Access: Examinee provides a word that has
a specific phonemic element in a specific location
(Sample items deleted for test security reasons)
32. COG Test 5B: Phonological ProcessingWord Fluency
• Word Fluency: Examinee names as many words as
possible that begin with a specified sound in 1 minute
(Sample items deleted for test security reasons)
33. COG Test 5B: Phonological ProcessingSubstitution
• Substitution: Examinee substitutes part of a word to
create a new word
(Sample items deleted for test security reasons)
35. Three-stage internal/ structural validity procedures for WJ IV battery
Stage 1:
Split-sample
random sample
generation
(total n = 7,416)
Model
Development
(MD)
samples (A)
Ages 3-5 A
(n = 209)
Model
Cross-Validation
(MCV)
samples (B)
Ages 3-5 B
(n = 208)
Stage 2:
Exploratory structural model
generation (MG) and evaluation
Stage 3:
Confirmatory
structural
model crossvalidation
phase
Stage 2 a
Exploratory structural
analysis in Sample A
Cluster analysis
(CA)
Model adjustment(s)
No
Ages 6-8 B
(n = 411)
+
Ages 9-13 A
(n = 785)
Ages 9-13 B
(n = 787)
Principal
component
analysis
(PCA)
Ages 14-19 A
(n = 842)
Ages 14-19 B
(n = 843)
Ages 20-39 A
(n = 625)
Ages 20-39 B
(n = 626)
Ages 40-90+ A
(n = 571)
Ages 40-90+ B
(n = 575 )
Ages 6-8 A
(n =412)
Specify initial
confirmatory
factor analysis
model (CFA)
+
Multidimensional
scaling analysis
(MDS)
Review of
contemporary
CHC and
cognitive
neuroscience
research
Review of
prior WJ,
WJ-R, WJ III
structural
validity
research
Copyright; Institute for Applied
Psychometrics; K.McGrew 1-23-14
Most
plausible
and best
fitting
odel?
CFA of MG model
result from MD (A)
sample in MCV (B)
sample(s)
Yes
36. Structural validity method comparisons across 7
intelligence batteries
EFA
Other exp.
methods
CFA
Cross-validation
WJ III
No
No
Yes
No
SB5
No
No
Yes
No
DAS-II
No
No
Yes
No
KABC-II
No
No
Yes (in “exploratory
manner”)
No
WAIS-IV
No
No
Yes
No
WISC-IV
Yes
No
Yes
Yes: Factor score (7)
congruence in random CV
sample of 440
WJ IV
Yes
Yes: MDS
and cluster
analysis
Yes: Exploratory
model generating
CFA (MD samples)
and CV CFA (CV
samples)
Yes: Randomly split
samples at six age groups.
CV of final MD models in
CV sample
Test
37. A
99.5 % of test
and latent
factor loadings
crossvalidated
(were
significant in
all agedifferentiated
samples crossvalidation B
samples)
!!!!!!
98.6 % of test
and latent
factor
loadings
crossvalidated
(were
significant in
all agedifferentiated
samples
crossvalidation B
samples) !!!!!!
B
Two final WJ IV cross-validated models across age groups: A = Broad CHC top-down model; B = Broad+narrow CHC bottom-up model
43. PHNPRO-Ga
SNDAWR-Ga
2.10
MDS relative CC (rCC)
Correlation = .93; but correspondence
diverges as test become higher in g and
cognitive complexity
VRBATN-Gwm
1.55
NUMSER-Gf
NUMREV-Gwm
SEGMNT-Ga
ORLVOC-Gc
ORLCMP-Gc
UNDDIR-Gwm
CONFRM-Gf
OBJNUM-Gwm
PICVOC-Gc
SENREP-Gwm
STYREC-Glr
1.00
RETFLU-Glr
LETPAT-Gs
NUMPAT-Gs
0.45
VISUAL-Gf
ANLSYN-GF
MEMWRD-Gwm
SNDBLN-Ga
NWDREP-Ga
PAIRCN-Gs
RPCNAM-Glr
PICREC-Gv
-0.10
0.4
0.5
g-loading
0.6
0.7
Relationship between g-loadings and MDS-based relative
cognitive complexity (rCC) for WJ IV COG and OL tests
0.8
44. Five primary design principles for WJ IV COG GIA
The WJ IV COG GIA cluster tests should:
1. Be the best factor indicators of each CHC broad domain
2. Be the best predictors of achievement from each CHC broad
domain
3. Be the most cognitively complex indicators from each CHC
broad domain
4. Be the best measures of g (general intelligence) from each
CHC broad domain
5. Collectively should have a relatively equal balance of type of
stimulus characteristics (verbal, numeric, figural)
Related
47. COG Test 2: Number Series
• Was in WJ III Diagnostic Supplement
• Gf-RQ (Quantitative Reasoning)
• Not a “controlled learning” test as are Concept
Formation (Gf-I) and Analysis-Synthesis (Gf-RG)
• More Gf “in the wild” – without examiner
provided scaffolding
• Extensive history as a premier Gf test in the
psychometric measurement of intelligence
• High in cognitive complexity and g. Best single test
predictor of achievement. Best indicator of Gf factor.
• In GIA, BIA, Gf-Gc Composite, Gf, Gf3, Quantitative
Reasoning (RQ), and one Math Aptitude clusters.
48. COG Test 2: Number Series
The examinee is
presented with a series
of numbers with one
number missing in the
series. The examinee
must determine the
missing number.
(Sample items deleted for test security reasons)
53. COG Test 3: Verbal Attention
• Measure of Gwm (working memory-WM; attentional
control-AC)
• More ecological “real world” valid measure of
working memory
• High in cognitive complexity and g. Within Gwm,
the most cognitively complex, one of best
indicators of Gwm factor, and best predictor
of achievement
• In GIA, BIA, Gwm, Gwm3, Cognitive Efficiency, and one
Reading and 1 Written Language Aptitude clusters.
54. COG Test 3: Verbal Attention
(Sample items deleted
for test security
reasons)
In addition to
working memory
(WM), task requires
attentional control
(AC: controlled
executive function;
aka, focus)
56. COG Test 4: Letter-Pattern
Matching
• Measure of Gs (perceptual speed) and orthographic
processing
• This speeded test (all WJ IV speeded tests) is based on a new
rate-based method of scaling the scores that eliminates the
need for bonus points
• Within Gs, it matches Number Pattern Matching in
g, Gs factor loading, and prediction of achievement. Is
more cognitively complex than Number Pattern Matching
• In GIA, Gs, Perceptual Speed (P), Cog. Eff. and Cog. clusters
57. COG Test 4: Letter-Pattern
Matching
(Sample items deleted
for test security
reasons)
Examinee locates
and circles the two
identical letter
patterns in a row of
six patterns in the
Response Booklet
60. OL Test 3: Segmentation
•
•
•
•
•
Ga (PC)
Examinee listens to words and identifies word parts
In OL Phonetic Coding (PC) cluster
Highest loading test on Ga factor across all ages
A moderate measure of g and predictor of ach. across
all ages; much more so (and more cognitively complex)
than Sound Blending.
• Such tasks have been reported to be strong predictors
of early reading (Bouwmeester et al, 2011; Geuden &
Sandra, 2003)
61. OL Test 3: Segmentation
(Sample items deleted for test security reasons)
63. COG Test 7: Visualization
• Measure of Gv-Visualization (Vz)
•Visualization consists of two subtests that each
measure Gv-Vz (visualization) via tasks that vary on
task complexity and degree of “minds eye” (mental
rotation) manipulations
• Within Gv, highest on cognitive complexity, g, Gv
factor, and prediction of achievement
• In GIA, Gv and both Math Aptitude clusters
64. COG Test 7A: Visualization-Spatial
Relations
(Sample items deleted for test security reasons)
Examinee identifies the two or three pieces that form a
complete target shape
65. COG Test 7B: Visualization-Block
Rotation
• Was in WJ III Diagnostic Supplement
• Gv-Vz (Visualization)
• Based on classic “mental rotation” tasks (e.g.,
Shepard& Metzlar; Vandenberg)
(Sample items deleted for test security reasons)
Examinee identifies the two block patterns that match
the target pattern, but have been rotated in space
66. COG Test 7B: Visualization-Block
Rotation
• Was in WJ III Diagnostic Supplement
• Gv-Vz (Visualization)
• Based on classic “mental rotation” tasks (e.g., Shepard
& Metzlar; Vandenberg)
68. WJ IV COG Scholastic Aptitude (SAPT) Clusters:
Primary design considerations
Back to the future: The SAPT concept is not new. SAPT’s were included in the WJ
and WJ-R; they were replaced with the Predicted Achievement option in the WJ III.
Each WJ IV COG SAPT cluster score is based on a combination of four tests that
together produce the most efficient and strongest prediction for the selected
achievement area.
The SAPTs are the combination of WJ IV COG tests that are the most highly
correlated with each curricular area as identified statistically (multiple regression);
consideration was also given to relevant theory and the extant research literature
(e.g., McGrew & Wendling,2010).
Some tests were eliminated as possible predictor tests (e.g., Number Series) if there
was clear predictor-criterion contamination (e.g., Math Problem Solving cluster
includes Number Matrices).
Each SAPT must include four tests that are each from different CHC broad ability
domains.
Copyright, Institute for Applied
Psychometrics, K.McGrew; 01-23-14
69. Clarification of Ability Construct Terminology
Copyright, Institute for Applied
Psychometrics, K.McGrew; 01-23-14
70. Ability
“as used to describe an attribute of individuals, ability refers to the possible
variations over individuals in the liminal levels of task difficulty (or in derived
measurements based on such liminal levels) at which, on any given occasion in
which all conditions appear favorable, individuals perform successfully on a
defined class of tasks” (p. 8, italics in original).
“every ability is defined in terms of some kind of performance, or potential for
performance (p. 4).”
Cognitive Abilities
Abilities on tasks “in which correct or appropriate processing of mental
information is critical to successful performance” (p. 10; italics in original).
Achievement abilities
“refers to the degree of learning in some procedure intended to produce
learning, such as an informal or informal course of instruction, or a period of
self study of a topic, or practice of a skill” (p. 17). As noted by Carroll (1993)
Copyright, Institute for Applied
Psychometrics, K.McGrew; 01-23-14
71. Abilities
Cognitive Abilities
Achievement Abilities
General
Intelligence (g)
Quantitative
Knowledge
(Gq)
Comp Knowledge
(Gc)
Fluid
Reasoning (Gf)
Short-Term
Memory (Gsm)
Long-Term
Storage &
Retrieval (Glr)
Visual
Processing
(Gv)
Auditory
Processing
(Ga)
Speed (Gs)
Etc.
Reading &
Writing (Grw)
Etc.
Etc.
Etc.
Etc.
Etc.
Etc.
Processing
Rdg
Apt
Math
Apt
Etc.
Etc.
Ach. domaingeneral apt.
Ach. domainspecific apt.
Copyright, Institute for Applied
Psychometrics, K.McGrew; 01-23-14
Vertical columns represent abilities, factors or latent traits
(primarily factor-analysis derived internal structural validity
constructs)
Horizontal arrow rows represent aptitudes (primarily multiple
regression derived external [predictive] validity constructs)
Conceptual distinction between Abilities: Cognitive abilities,
achievement abilities, and aptitudes (McGrew, in press b)
72. WJ IV Scholastic Aptitude Cluster Organization
Gf
Concept
Formation
(I)
Gc
Gwm
Oral
Vocabulary
(LD/VL)
Oral
Vocabulary
(LV/VL)
Number
Series
(RQ)
Oral
Vocabulary
(LD/VL)
Verbal
Attention
(WM)
Num Pattern
Matching
(P)
Num Pattern
Matching
(P)
WJ IV Ach Clusters
Reading
Broad Reading
Reading Comp
Reading Comp-Ext
Reading Fluency
Reading Rate
Phonological
Processing
(PC)
Basic Rdg Skills
Writing
Broad Writing
Written Expression
Story
Recall
(MM)
Basic Writing Skills
Visualization
(Vz)
Math
Broad Math
Math Calc Skills
Visualization
(Vz)
Pair
Cancellation
(P/EF)
Numbers
Reversed
(WM)
Gv
Phonological
Processing
(PC)
Phonological
Processing
(PC)
Glr
Phonological
Processing
(PC)
Num Pattern
Matching
(P)
Verbal
Attention
(WM)
Oral
Vocabulary
(LD/VL)
AnalysisSynthesis
(RG/RQ)
Ga
Num Pattern
Matching
(P)
Oral
Vocabulary
(LV/VL)
Oral
Vocabulary
(LD/VL)
Gs
Math Prob Solving
Grw/Gq domain general
Grw domain specific
Gq domain specific
Copyright, Institute for Applied
Psychometrics, K.McGrew; 01-23-14
73. Scholastic Aptitude Clusters: Potential Uses
•
•
•
•
•
•
Designed to predict near term academic performance
Time efficient referral-focused selective testing
Time efficient academic domain-specific screening
Time efficient annual review evaluations
Gifted and talented screening – domain-specific talents?
May be useful in estimating quickness of response to
intervention
• Provide information regarding the concept of “expected
underachievement”
• Formulation of differential academic domain expectations
• ?
74. CHC achievement
abilities and WJ IV clusters
CHC cognitive abilities and WJ IV scholastic aptitude cognitive clusters
Gf
Gc
Gwm
Gs
Ga
Glr
Gv
Gq
Grw
Reading
Rdg. Cmp.
Rdg. Apt. A
(Predictor Score)
Regression-based prediction models that
account for regression-to-the-mean
Brd. Rdg.
Rdg. Flu.
Rdg Cmp. Ex
Rdg. Rate
Rdg. Apt. B
(Predictor Score)
Bas. Rdg. Sk.
Wr. Lng.
WL. Apt. A
(Predictor Score)
Brd. Wr. Lg.
Wr. Exp.
WL. Apt. B
(Predictor Score)
Bas. Wr. Sk.
Math.
Math Apt. A
(Predictor Score)
Brd. Math
Math Cal.Sk.
Math Apt. B
(Predictor Score)
Predicted
Target Cluster
Score
Math Pr. Slv.
(minus)
(See Table 1-10 for the specific tests and
cluster information)
Scholastic Aptitude/Achievement comparison procedures
Copyright, Institute for Applied
Psychometrics, K.McGrew; 01-23-14
Actual Target
Cluster Score
Difference
Score
(equals)
(Compare to distribution of difference
scores in WJ IV norm sample to
determine significant strength or
weakness)
SD and PR for specified
SD cut-off score
75. Additional select technical and
psychometric information
(McGrew, LaForte, Schrank, 2014)
Dr. Kevin McGrew
Institute for Applied Psychometrics (IAP)
Copyright; Institute for Applied
Psychometrics; K.McGrew 02-5-14
76. National Norm Sample
• 7,416 participants
• Preschool (664)
• K-12 (3,891)
• College/University (775)
• Adult (2,086)
• Ages 2-90+ years, Grades K.0-18.0
• 100 geographically diverse communities from 46 states and the
District of Columbia
• The Standards for Educational and Psychological Testing (aka., the
Joint Standards) (American Psychological Association, American
Educational Research Association, & National Council on
Measurement in Education, 1999; in press) guided the norming
and technical analysesCopyright; Institute for Applied
Psychometrics; K.McGrew 1-23-14
77. Sources of validity
evidence presented
•Representativeness of the WJ IV Test Content, Process, and
Construct Coverage
•Developmental Patterns of WJ IV Ability Clusters
•Internal Structure and Relations within the WJ IV
•Relationship of WJ IV Scores to Other Measures of Cognitive
Abilities, Oral Language, and Achievement
•Performance of Clinical Samples on WJ IV Measures
Copyright; Institute for Applied
Psychometrics; K.McGrew 02-5-14