1 / 54

FACTOR ANALYSIS

FACTOR ANALYSIS. LECTURE 11 EPSY 625. PURPOSES. SUPPORT VALIDITY OF TEST SCALE WITH RESPECT TO UNDERLYING TRAITS (FACTORS) EFA - EXPLORE/UNDERSTAND UNDERLYING FACTORS FOR A TEST CFA - CONFIRM THEORETICAL STRUCTURE IN A TEST. HISTORICAL DEVELOPMENT.

gavril
Télécharger la présentation

FACTOR ANALYSIS

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. FACTOR ANALYSIS LECTURE 11 EPSY 625

  2. PURPOSES • SUPPORT VALIDITY OF TEST SCALE WITH RESPECT TO UNDERLYING TRAITS (FACTORS) • EFA- EXPLORE/UNDERSTAND UNDERLYING FACTORS FOR A TEST • CFA- CONFIRM THEORETICAL STRUCTURE IN A TEST

  3. HISTORICAL DEVELOPMENT • PEARSON (1901)- eigenvalue/eigenvector problem (dimensional reduction) “method of principal axes) • SPEARMAN (1904) “General Intelligence, Objectively Measured and Determined” • Others: Burt, Thompson, Garnett, Holzinger, Harmon, Thurstone

  4. FACTOR MODELS SUBJECTS FIXED SAMPLE Fixed VARIABLES Principal components, common factors Image Sampl e ALPHA Factor Analysis Canonical Factor Analysis

  5. EXPLORATORY FACTOR ANALYSIS • USE PRINCIPAL AXIS METHOD: • ASSUMES THERE ARE 3 VARIANCE COMPONENTS IN EACH ITEM: • COMMONALITY (h2) • UNIQUENESS: • SPECIFICITY (s2) • ERROR (e2)

  6. SINGLE FACTOR • REQUIRES AT LEAST 3 ITEMS OR MEASUREMENTS TO UNIQUELY DETERMINE

  7. ASSUMED=0 FOR PARALLEL ITEMS SPECIFICITY CALLED FACTOR LOADING CORRELATION BETWEEN ITEM AND FACTOR .714 e ITEM1 .7 .6 e FACTOR ITEM2 .8 .6 .8 ITEM3 e

  8. ASSUMED=0 FOR PARALLEL ITEMS ALPHA= SPEARMAN-BROWN STEPPED UP AVERAGE INTER ITEM CORRELATION: (.56 +.42+.48)/3=.49 ALPHA= 3(.49)/[1+2(.49)] = .74 SPECIFICITY .714 e ITEM1 .7 =1-.72 .6 e FACTOR ITEM2 .8 .6 .8 ITEM3 e

  9. TWO FACTORS • NEED AT LEAST 2 ITEMS OR MEASUREMENTS PER FACTOR, ASSUMING FACTORS ARE CORRELATED

  10. .7 e ITEM1 FACTOR 1 e ITEM2 ITEM 3 e .8 .6 e .5 CORRELATION BETWEEN FACTORS FACTOR 2 ITEM 4 .7

  11. CORRELATION BETWEEN ANY TWO ITEMS = PRODUCT OF ALL PATHS BETWEEN THEM; EX. R(ITEM1, ITEM4) = .7 x .5 x .7 = .245 .7 e ITEM1 FACTOR 1 e ITEM2 ITEM 3 e .8 .6 e .5 CORRELATION BETWEEN FACTORS FACTOR 2 ITEM 4 .7

  12. SIMPLE STRUCTURE • TRY TO CREATE SCALE IN WHICH EACH ITEM CORRELATES WITH ONLY ONE FACTOR: ITEM FACTOR 1 2 3 ITEM 1 1 0 0 ITEM 2 1 0 0 ITEM 3 0 1 0 ETC

  13. CRITERIA FOR SIMPLE STRUCTURE • Structural equation modeling provides chi square test of fit • Compares observed covariance (correlation) matrix with predicted/fitted matrix • Alternatively, look at RMSEA (Root mean square error of approximation) of deviations from fitted matrix

  14. MATHEMATICAL MODEL • Z = persons by variables matrix of p x k standardized variables (mean=0, SD=1) • Z’Z = NR (covariance matrix) k x k • Zi = aiFi + ei

  15. MATHEMATICAL MODEL • Z = AF = C + U • ZZ’/N = R = AFF’A’ + U2 • S = ZF’/N (structure matrix: correlations between Z and F) = AFF’/N •  = FF’/N (correlations among factors) • A = Pattern matrix

  16. MATHEMATICAL MODEL • S = A • A = S -1 (If factors uncorrelated, A=S) Pattern matrix = Structure matrix • R = ZZ’/N = CC’/N + U2

  17. MATHEMATICAL MODEL • If we take the covariance matrix of F to be diagonal, and the metric of variances of Fi to be 1.0, • R = AA’/N = SA’ = AS’

  18. MATHEMATICAL MODEL • Now let Zi = aiFi + si + ei • Let Ŕ = R - D2, where D2 is a diagonal matrix of specificities and error: si + e2i • Then Ŕ = AFF’A/N = A A’ = SA’ = AS’ •  = I  Ŕ = AA’

  19. MATHEMATICAL MODEL • How do we estimate s2i ? • Instead, estimate [R2- U2]ii= [I- s2i - e2i]ii • Consider for each zi that it is predictable from the rest: zi = b1z1 + b2z2 + …bi-1zi-1 + ... • Then R2i = variance common to all other variables (squared multiple correlation or SMC)  h2i = communality for item i • Due to Dwyer (1939)

  20. MATHEMATICAL MODEL • SMC is estimable from the observed data, so that Ŕ = R - [1-SMCi] where [SMCi] = diagonal matrix with SMCs for each variable on the diagonals and zeros off-diagonal • Theorem states “SMCs guarantee that the number of factors  # eigenvalues>1.0

  21. MATHEMATICAL MODEL Ŕ = R21.234.. 0 0 0 0 … 0 R22.134.. 0 0 0 … 0 0 R23.124.. 0 0 … 0 0 0 R24.123.. 0 …

  22. MATHEMATICAL MODEL • SOLUTIONS: PRINCIPAL COMPONENTS (R = Ŕ ) Rq = q, RQ = Q, = diagonal [i] Q-1RQ =  QQ’ = I = Q-1 = Q’ Q’RQ =  (Spectral Theorem)

  23. MATHEMATICAL MODEL • SOLUTIONS: PRINCIPAL AXIS ( Ŕ- I)q = 0 • That is, solve for first eigenvalue • | Ŕ- I | = 0, solved by • Rmq = mq begin with m=2: R2q = 2q , then put solution in R(Rq1) = 2q1, iterate for m=4

  24. MATHEMATICAL MODEL • Now compute residual correlation matrix: R21 = R2 -  Ŕ , iterate

  25. EIGENVALUES • i = variance of ith factor • i / i = proportion of total variance accounted for by the ith factor • i < 1  chance factor • Scree plot (value x factor eigenvalue ordered from greatest to lowest)

  26. K 1.0 0 SCREE PLOT 1 2 3 4 5 6 7 . . . . K

  27. ROTATION • MEANING CRITERION: SIMPLE STRUCTURE POSITIVE MANIFOLD • B=AT A=INITIAL FACTOR MATRIX T=TRIANGULAR MATRIX B=FINAL FACTOR MATRIX • TT’=

  28. VARIMAX ROTATION (uncorrelated Factors) • ORTHOGONAL (RIGID) ROTATION • Maximize V=n  (bjp/hj)4 - (b2jp/h2j)2 • Geometric problem: • (X,Y) = (x,y) cos - sin  sin  - cos 

  29. VARIMAX ROTATION • (X,Y) = (x,y) cos - sin  • sin  - cos  uj = x2j - y2j vj = 2xjyj A= uj B= vj C= (uj - vj)2 D=2 ujvj solve tan4 = [D-2AB/h]/[C-(A2-B2)/h] -45o   45o

  30. Orthogonal (perpendicular) Rotation of Axes  Unrotated Factor 1 loading values Unrotated Factor 2 loading values

  31. OBLIQUE SOLUTION (correlated Factors) • MINIMIZE S (OBLIMIN) • S =  [n(v2jp/h2j)(v2jg/h2j)- ((v2jp/h2j)((v2jg/h2j)] • PROMAX: • Start with VARIMAX, B=AT, transform with • vjp = (bjp4)/bjp

  32. FACTOR CORRELATION •  = TT’ • Tij = cos(ij) -sin(ij) sin(ij) cos(ij) • rij= [cos(ij)(-sin(ij)]+ [sin(ij)cos(ij)] = T11T12 + T21T22

  33. FACTOR CORRELATION • S = P  (Structure matrix= Pattern matrix x factor correlation matrix) • P = A(T’)-1 • A = PT’

  34. Oblique Rotation of Axes ij 

  35. ALPHA FACTOR ANALYSIS • Estimates population h2i for each variable • Little different from common factors

  36. Canonical Factor Analysis • Uses canonical analysis to maximize R between factors and variables, iterative Maximum Likelihood analysis

  37. Image Analysis • h2i = R2i.1,2,…K • pj = wjkzk (standard regression) • ej = zj - pj called anti-image • Var(ej)> Var(j) where Var(j) = anti-image for the regression of zj on the factors F1,F2, …FK

  38. FACTOR CONGRUENCE • Alternative to Confirmatory Analysis for two groups who it is hypothesized have the same factor structure: • Spq = ajpbjq / [a2jp b2jq ] • This is basically the correlation between factor loadings on the comparable factors for two groups

  39. Example of 2 factor structure • Achievement (reading, math) and IQ (verbal, nonverbal) • quasi-multitrait multimethod analysis: • reading is verbal • math is “nonverbal”

  40. CONFIRMATORY FACTOR ANALYSIS

  41. BASIC PRINCIPLES • xx´) • 2x1 • xx = x1x2 2x2 • x1x3 x2x3 2x3

  42. BASIC PRINCIPLES • 2x1 =2111 + 21 • 2xk =2k11 + 2k • xixk =x111 xk 1 x1 1 xk k

  43. IDENTIFICATION RULES • t-rule : tq(q+1), q=#manifest variables • necessary but not sufficient • 3-indicator rule: 1 factor3 indicators • sufficient but not necessary • 2-indicator rule: 2+ factors2 indicators @ • local vs. global identification: • local: sample estimates of parameters independent- necessary but not sufficient • global: population parameters independent- necessary and sufficient

  44. ESTIMATION • MODEL EVALUATION • FIT: FML used to evaluate , S • Residuals: E= S -  • RMR = SD(sij - ij ) • RMSEA = √[(2/df - 1) /(N - 1)] • note: factor analyze E, should be 0 ˆ ˆ ˆ ˆ ˆ

  45. Hancock’s Formula- reliability for a given factor Hj = 1/ [ 1 + {1 / (Σ[l2ij/(1- l2ij )] ) } Ex. l1 = .7, l2= .8, l3 = .6 H = 1 / [ 1 +1/( .49/.51 + .64/.36 + .36/.64 )] = 1 / [ 1 + 1/ ( .98 +1.67 + .56 ) ] = 1/ (1 + 1/3.21) = .76

  46. Hancock’s Formula Explained Hj = 1/ [ 1 + {1 / (Σ[l2ij/(1- l2ij )] ) } now assume strict parallelism: then l2ij= 2xt thus Hj = 1/ [ 1 + {1 / (Σ[2xt /(1- 2xt)] ) } = k 2xt / [1 + (k-1) 2xt ] = Spearman-Brown formula

  47. TEST • (n-1)FML ~ t • used for nested model: model with one or more restrictions from original • restriction = known parameter, equality of two or more parameters • Proof: Bollen shows (N-1)[-2Log(L0/L1)= (N-1)FML where L0 is unrestricted, L1 restricted models

  48. INCREMENTAL FIT • Bentler and Bonnet: 1 = Fb - Fm Fb = b - m b • can be used to compare improvements over original model or against a standard or baseline

More Related