PPSYDR
,4. TEST DIMENSIONALITY AND FACTOR ANALYSIS
3DIMEUSIORALITY QUESTOUS:
UNIDIMENSIONAL TESTS
the testincludes items thatreflect
When
only a
single attribute of a
person
->
the
responses to the testitems are driven
only by thatone attribute)
test
eg. abt.geometry which
requires true
and ONLY tests
knowledge of geometry
geometry (ne
angebra or stn.)
MULTIDIMENSIONAL TESTS W/
CORRELATED DIMENSIONS
ana tests w/ higher order factors
when testitems reflect more than
one
psychological attribute +
the dimensions correlated
if are
->
only a
single score is computed eg. intelligence tests w/ questions abt
diff facets of intelligence (logic
thinking general knowledge...)
when dimensions are carrelated:person
wI
high score in edimension is
likely
to score
high on other dimension
typically
->
produce more scores,
each subset has its own score
-> when correlated: dimension are
scared to produce a total score by
subsets
combining
, MULTDIMENSIONAL TESTS N/ UnCORRELATED DIMEnSIONS
eg. some personality tests which testfive relatively independent
personcuity attributes
->
test receives five scores which are all treated as if they
were
unidimensional, no total testscore is computed
FACTOR ARALYSIS
this statistical procedure
evaluates the test dimension -
ality
there are two
types:
EXPLORATORY FACTOR ANALYSIS GFA
compute correlations among items and
identify how many dimensions
there really are:
in the
hypothetical correlation matrix
we look for items thatare strongly
correlated w/ each other not
but
other items
=psychological dimension/factor
tallative, assertive, outgoing
-
->
imaginative, creative, intellectual
, EFA STEPS:
1. EXTRACTION METHOD
CHOOSE
which technique? PRINCIPAL AXIS FACTORING (PAF) recommended*
PRIACIPAL COMPONENTS ANALLSIS (PCA)
MAX. LIKELIHOOD FACTOR ANALYSIS
2.IDEUTFY #OF FACTORS +
EXTRACTION
EIGENVALUES
-to
identify # of factors look at
each value represents the potential #of
dimensions reflected among test items
OPTON ONE:
-
than 1.0 then reflects
when itis
greater the item a dimension
eg. if two are 31 then the items reflect I dimensions
OPMON TO:
-
examine relative
the sizes
of the
figenvalues
the 2 and 3
biggestrel, diff. is btwn value
we find the point at which subsequent
differences are relative small (So point3)
and we conclude the test has that
point- 1 dimensions (So 3-1 2)=
we can also plot these values in a SCREf PLOT
wherever the
"leveling off-point"is, then-1
3 1 2D
-
=
D
,4. TEST DIMENSIONALITY AND FACTOR ANALYSIS
3DIMEUSIORALITY QUESTOUS:
UNIDIMENSIONAL TESTS
the testincludes items thatreflect
When
only a
single attribute of a
person
->
the
responses to the testitems are driven
only by thatone attribute)
test
eg. abt.geometry which
requires true
and ONLY tests
knowledge of geometry
geometry (ne
angebra or stn.)
MULTIDIMENSIONAL TESTS W/
CORRELATED DIMENSIONS
ana tests w/ higher order factors
when testitems reflect more than
one
psychological attribute +
the dimensions correlated
if are
->
only a
single score is computed eg. intelligence tests w/ questions abt
diff facets of intelligence (logic
thinking general knowledge...)
when dimensions are carrelated:person
wI
high score in edimension is
likely
to score
high on other dimension
typically
->
produce more scores,
each subset has its own score
-> when correlated: dimension are
scared to produce a total score by
subsets
combining
, MULTDIMENSIONAL TESTS N/ UnCORRELATED DIMEnSIONS
eg. some personality tests which testfive relatively independent
personcuity attributes
->
test receives five scores which are all treated as if they
were
unidimensional, no total testscore is computed
FACTOR ARALYSIS
this statistical procedure
evaluates the test dimension -
ality
there are two
types:
EXPLORATORY FACTOR ANALYSIS GFA
compute correlations among items and
identify how many dimensions
there really are:
in the
hypothetical correlation matrix
we look for items thatare strongly
correlated w/ each other not
but
other items
=psychological dimension/factor
tallative, assertive, outgoing
-
->
imaginative, creative, intellectual
, EFA STEPS:
1. EXTRACTION METHOD
CHOOSE
which technique? PRINCIPAL AXIS FACTORING (PAF) recommended*
PRIACIPAL COMPONENTS ANALLSIS (PCA)
MAX. LIKELIHOOD FACTOR ANALYSIS
2.IDEUTFY #OF FACTORS +
EXTRACTION
EIGENVALUES
-to
identify # of factors look at
each value represents the potential #of
dimensions reflected among test items
OPTON ONE:
-
than 1.0 then reflects
when itis
greater the item a dimension
eg. if two are 31 then the items reflect I dimensions
OPMON TO:
-
examine relative
the sizes
of the
figenvalues
the 2 and 3
biggestrel, diff. is btwn value
we find the point at which subsequent
differences are relative small (So point3)
and we conclude the test has that
point- 1 dimensions (So 3-1 2)=
we can also plot these values in a SCREf PLOT
wherever the
"leveling off-point"is, then-1
3 1 2D
-
=
D
