Exploratory factor analysis Flashcards
Exploratory factor analysis and principle component analysis are hypothesis ________ techniques rather than hypothesis ________ techniques.
hypothesis generating rather than hypothesis testing
X refers to the observed variables, which are _____________________________
measured variables - i.e. questionnaire or physiological measures
Y refers the latent variables, which are ________________________________
personality factors ( extraversion, neuroticism)or physiological area of interest
Y is sort of a measure of X
U represents the ________ of ______ variables (Y) onto _______ variables (X)
U= projection latent variables onto observed variables
Latent variables are a ______ of observed variables
X = __ x____
X = U x Y
X = observed, x = multiply
Transform, or __, is the inverse of __.
Transform = V, which is the inverse of U
Y = X times __
A V
B U
C Y
D X
A , Y = X times V
V = u-1, the ________ of U
V is the inverse of U
V is the _________ matrix, accomplishing the projection. Vectors contributing to this matrix are called ___________
V is the rotation matrix
Vectors contributing to rotation = eigenvectors
(extracted variables are called principle components)
What are the 2 constraints of principle component analysis?
What do these constraints ensure
- Principle components successively explain maximum variance - ensuring as much info as possible is captured in as few variables as possible
- Principle components are mutually orthogonal (uncorrelated) -ensures no redundancy
(Comparing PEN vs RST)
Factors can be observed in a ___________ system.
coordinate
A PCA/EFA can measure correlations of items ________ a __________, and correlations of items _________ ______
Measures correlations of items within a group - want this to be high
Measures correlations of items between groups - want this to be low
The cut off where eigenvalues fall approximately directly down is called the
A Infraction point
B Reflection point
C Inflection point
D Infection point
C inflection point
What are two possible criteria for the cut off of values
inflection point
When factors provide less contribution than an original single factor
A variables _______ ________ determines its position in the coordinate system
factor loadings determine position in the coordinate system.
The factor loadings help reveal the meanings of the ______ variables, and together with ___________ can reveal the hidden structure in a large data set.
Factor loadings help reveal meanings of hidden variables
with eignenvectors can reveal structure in a data set.
Factor analysis goes a step further than PCA, showing factor loadings of each _________, and adding a ________
Factor analysis shows factor loadings of each variable and adds an additional rotation
Principle component analysis is a good _____ ______ for factor analysis, transforming raw data into a __________ system.
PCA a good starting point for factor analysis
transforms raw data into coordinate system
PCA’s only two concerns are _________________________________ and __________, it does not care about how distinctively _____________ each PC is.
PCA only concerned in explaining maximum variance and orthogonality
doesn’t care about how distinctively interpretable each pc is
The rotation from PCA to factor analysis is called a _____________ rotation.
varimax
Factor analysis makes factor loadings more ________.
Ideally we want ______ factor loadings (Positive or negative) of variables within their factors, and _______ (close to 1) factor laodings of variables with other factors. We do not want ___________.
Factor analysis makes factor loadings more distinct
we want high factor loadings of variables within their factors - close to 1 or -1
and low factor loadings of variables with other factors - no overlap.
Varimax rotations are ____________.
Oblimin/promax rotations are ____________, losing __________.
varimax rotations are orthogonal
oblimin/promax rotations are oblique, losing orthogonality
If the bartlett’s test of sphericity is signifcant, the data is not ___________, so we ________ run a correlation matrix. If the
bartlett’s test of data is significant, the data is not spherical
so we should run a significance matrix
If data is spherical, should not run a matrix
Items scored negatively have ________ factor laodings to their factor, slose to ___.
Items scored positively have _________ factor loadings to their factor, close to __.
Negative scores have negative factor loadings close to -1
Positive scores have positive factor loadings close to 1
The Kaiser Mayer Olkin test tests for
A sphericity
B normality
C sampling adequacy
D orthogonality
C sampling adequacy
What is a good result in a Kaiser Mayer Olkin test, and what is a bad result?
good result = 0.9-1
bad result = 0.0-0.5
The ratio of participants over items should be between : and : and the min number of participants should be ______.
should be between 5:1 and 10:1
min 100 participants
Communalities are masured by a ________ value
If communalities are more than 0.6, a minimum of ____ participants are needed.
If communalities = 0.5 and there are few factrors, the number of participants should be between ____ and _____
If communalities are less than 0.5, and there are many factors, a minimum of _____ participants are needed
Communalities = hsquared value
If communalities above 0.6, need min of 100 participants
if communalities = 0.5 and few factors, need 100-200 participants
If communalities are below 0.5 and many factors, need above 500 participants
To measure internal consistency, we can use a _________ _______ test or a _______-________ __ test.
To measure external reliability, test-retest reliability should be more than ____
Internal consistency test can use Cronbach’s alpha or Kuder Richardson -20 test
test retest reliability rates should be above 0.7
Reliability and internal consistency refer to the quality of the __________, rather than the _______ ______.
relaibility and internal consistency refer to quality of questionnaires rather than factor analysis
