Title: Common Factor Analysis
1Common Factor Analysis
- World View of PC vs. CF
- Choosing between PC and CF
- PAF -- most common kind of CF
- Communality Communality Estimation
- Common Factor Scores
2World View of PC Analyses
- PC analysis is based on a very simple world
view - We measure variables
- The goal of factoring is data reduction
- determine the of kinds of information in the
variables - build a PC for each
- R holds the relationships between the variables
- PCs are composite variables computed from linear
combinations of the measured variables
3World View of CF Analyses
- CF is based on a somewhat more complicated and
causal world view - Any domain (e.g., intelligence, personality) has
some set of latent constructs - A persons values on these latent constructs
causes their scores on any measured variable(s) - any variable has two parts
- common part -- caused by values of the latent
constructs - unique part -- not related to any latent
construct (error)
4World View of CF Analyses, cont
- the goal of factoring is to reveal the number and
identify of these latent constructs - R must be adjusted to represent the
relationships between portions of the variables
that are produced by the latent constructs - represent the correlations between the common
parts of the variables - CFs are linear combinations of the common parts
of the measured variables that capture the
underlying constructs
5Example of CF world view
- latent constructs
- IQ Math Ability Reading Skill Social
Skills - measures
- adding, subtraction, multiplication
vocabulary, reading speed,
reading comprehension politeness, listening
skills, sharing skills - Each measure is produced by a weighted
combination of the latent constructs, plus
something unique to that measure . . . - adding .5IQ .8Math 0Reading 0Social
Ua - subtraction .5IQ .8Math 0Reading
0Social Us - vocabulary .5IQ 0Math .8Reading
0Social Uv - politeness .4IQ 0Math 0Reading
.8Social Up
6Example of CF world view, cont
- When we factor these, we might find something
like - CF1 CF2 CF3 CF4
- adding .4 .6
- subtraction .4 .6
- multiplication .4 .6
- vocabulary .4 .6
- reading speed .4 .6
- reading comp .4 .6
- politeness .3 .6
- listening skills .3 .6
- sharing skills .3 .6
- Name each latent construct that was revealed by
this analysis
7Principal Axis Analysis
- Principal again refers to the extraction
process - each successive factor is orthogonal and accounts
for the maximum available covariance among the
variables - Axis tells us that the factors are extracted
from a reduced correlation matrix - diagonals lt 1.00
- diagonals the estimated communality of each
variable - reflecting that not all of the variance of that
variable is produced by the set of latent
variables - So, factors extracted from the reduced R will
reveal the latent variables
8Which model to choose -- PC or PAF
? Traditionally...
- PC is used for psychometric purposes
- reduction of collinear predictor sets
- examination of the structure of scoring systems
- consideration of scales and sub-scales
- works with full R because composites will be
computed from original variable scores not
common parts - CF is used for theoretical purposes
- identification of underlying constructs
- number and identity of basic elements of
behavior - The basis for latent class analyses of many
kinds - both measurement structural models
- works with reduced R because it hold the
meaningful part of the variables and their
interrelationships - The researcher selects the procedure based on
their purpose for the factor analysis !!
9Communality Its Estimation
- The communality of a variable is the proportion
of that variables variance that is produced by
the common factors underlying the set of
variables - Common Estimations
- ? (reliability coefficient) -- only the reliable
part of the variable can be common - largest r (or r2) with another in the set -- at
least that much is shared with other variables - R2 predicting that variable from all the others
-- tells how much is shared with other variables - Note how the definition shifts from variance
shared with the latent constructs to variance
shared with the other variables in the set !!
10Communality Its Estimation How SPSS does it
- Step 1 Perform a PC analysis
- extract PCs from the full R matrix
- Step 2 Perform 1st PAF Iteration
- Use R2 predicting each variable from others --
put in diagonal of R - extract same PAFs from that reduced R matrix
- compute (output) variable communalities
- Step 3 Perform 2nd PAF Iteration
- use variable (output) communalities from last PAF
step as estimated (input) communalities -- put in
diagonals of R - extract same PAFs from that reduced R matrix
- compute (output) variable communalities
- Compare estimated (input) and computed (output)
variable communalities - Additional Steps Iterate to convergence of
estimated (input) computed
(output) variable communalities
11Communality Its Estimation How SPSS does it,
cont
- Huh?!!?
- The idea is pretty simple (and elegant)
- If the communality estimates are correct, then
they will be returned from the factor analysis ! - So, start with a best guess of the
communalities, and iterate until the estimates
are stable - Note This takes advantages of the
self-correcting nature of this iterative
process - the initial estimates have very little effect on
the final communalities (R2 really easy to
calculate) - starting with the PC communalities tends to work
quickly - Note This process assumes the latent constructs
are adequately represented by the variable set !!
12Problems estimating communalities in a CF analysis
- failure to converge
- usually this can be solved by increasing the
number of iterations allowed (1000) - Heywood case ? ? gt 1.00
- During iteration communality estimates can
become larger than 1.00 - However no more than all of a variables
variance can be common variance! - Usual solutions
- Use the solution from the previous iteration
- Drop the offending variable
- If other variables are threatening to Heywood
consider aggregating them together into a single
variable
13Common Factor Scores
- The problem is that common factors can only be
computed as combinations of the common parts of
the variables - Unfortunately, we cant separate each persons
score on each variable into the common and
unique part - So, common factor scores have to be estimated
- Good news --
- the procedure used by SPSS works well and is well
accepted - since CF is done for theory testing or to
reveal latent constructs rather than for
psychometric purposes scores for CFs are not
used as often as are PC scores
14Maximum Likelihood method of Common Factoring
- Both PAF ML are common factor extractions
- they both seek to separate the common vs.
unique portion of each variables variance and
include only the common in R - they both require communality estimates
- they both iterate communality input estimates
output computations until these two converge,
though the process for computing estimates is
somewhat different - which is taken as evidence that the communality
estimates are accurate and so, S extracted using
those estimates describes the factor structure of
R - PAF factors are extracted to derive S that will
give the best reproduction of variance in
sampled R matrix - ML factors are extracted to derive S that is
most likely to represent population S
reproduce the population R
15Maximum Likelihood method of Common Factoring
- If assumptions of interval measurement and
normal distribution are well-met, ML works
somewhat better than PAF vice versa - ML is an extraction technique the rotational
techniques discussed for PC and PAF all apply to
ML factors - ML is a common factoring technique issue of
factor score estimation are the same as for
PAF - Proponents of ML exploratory factoring emphasize
- ML estimation procedures are most the common in
confirmatory factoring, latent class measurement,
structural models the generalized linear model - ML estimation permits an internally consistent
set of significance tests e.g., factors
decisions.