lecture17_s10 EFA

# lecture17_s10 EFA - Exploratory Factor Analysis: common...

This preview shows pages 1–6. Sign up to view the full content.

Exploratory Factor Analysis: common factors, principal components, and more Psychology 588: Covariance structure and factor models Apr 14, 2010

This preview has intentionally blurred sections. Sign up to view the full version.

View Full Document
• As learned earlier for CFA, both CFA and EFA are of (almost) the same model form, which defines q indicator variables as linear combinations of n “common” factors plus q mutually orthogonal “unique” factors as follows: Common factor model 2 ( ) () 111 ,, 0 nq i j qnq E Ei j EE E δδ × ×× × =+ = = === x Λξ δ ξδ 0 x0 ξ 0 δ 0 for Given • Difference between CFA and EFA resides in whether some selective elements of loading matrix Λ are constrained at particular constants (e.g., at 0 or by equality) --- consequently, the parameters are estimated by different methods
• Technically, CFA and EFA differ by the degree of constraints ¾ EFA --- loading matrix has minimal constraints for a unique solution with fixed orientation of factors To fix orientation of factor axes, n × n “constraints” need be imposed, and that is done by n scaling constraints, n ( n –1)/2 elements for orthogonal factors ξ and n ( n elements for the loading matrix Λ (i.e., canonical form) ¾ CFA --- further constraints (motivated by theoretical hypotheses) imposed to see if they agree with the data • Furthermore, CFA may allow correlated measurement errors (so long as all parameters identifiable), but EFA doesn’t allow such relaxation by definition

This preview has intentionally blurred sections. Sign up to view the full version.

View Full Document
() ( )( ) ( ) 12 , diag , ,..., q E EE E E θθ θ ′′ ≡= + + = =+ ≡≡ Σ xx Λξ δ ξ Λ δ x0 ΛΦΛ Θ Φξ ξ Θδ δ Implied cov matrix of the common factor model 4 • It’s a convention in EFA to scale the common factors to have a variance of 1 (instead of setting their metric equal to one of their indicators) --- by this scaling, Φ becomes a correlation matrix • When considered for realized data (i.e., “subjects”), ξ is often called “factor score” matrix
• Common factors ξ are constrained to be mutually orthogonal, and so with the unit-variance scaling, factors becomes mutually orthogonal z-scores: Orthogonal factor model 5 • Thanks to the rotational indeterminacy, orthogonal common factors are typically estimated in a canonical form for a

This preview has intentionally blurred sections. Sign up to view the full version.

View Full Document
This is the end of the preview. Sign up to access the rest of the document.

## This note was uploaded on 06/11/2010 for the course PSYC 588 taught by Professor Sunjinghong during the Spring '10 term at University of Illinois at Urbana–Champaign.

### Page1 / 17

lecture17_s10 EFA - Exploratory Factor Analysis: common...

This preview shows document pages 1 - 6. Sign up to view the full document.

View Full Document
Ask a homework question - tutors are online