1 / 60

Leuven Workshop August 2008

The Measurement and Analysis of Complex Traits Everything you didn’t want to know about measuring behavioral and psychological constructs. Leuven Workshop August 2008. Overview. SEM factor model basics Group differences: - practical Relative merits of factor scores & sum scores

foy
Download Presentation

Leuven Workshop August 2008

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. The Measurement and Analysis of Complex TraitsEverything you didn’t want to know about measuring behavioral and psychological constructs Leuven Workshop August 2008

  2. Overview • SEM factor model basics • Group differences: - practical • Relative merits of factor scores & sum scores • Test for normal distribution of factor • Alternatives to the factor model • Extensions for multivariate linkage & association

  3. Structural Equation Model basics • Two kinds of relationships • Linear regression X -> Y single-headed • Unspecified Covariance X<->Y double-headed • Four kinds of variable • Squares – observed variables • Circles – latent, not observed variables • Triangles – constant (zero variance) for specifying means • Diamonds -- observed variables used as moderators (on paths)

  4. 1.00 F lm l1 l2 l3 S1 S2 S3 Sm e1 e2 e3 e4 Single Factor Model

  5. Factor Model with Means MF 1.00 1.00 mF1 mF2 F1 F2 lm l1 l2 l3 S1 S2 S3 Sm mSm e1 e2 e3 e4 mS2 mS3 mS1 B8

  6. Factor model essentials • Diagram translates directly to algebraic formulae • Factor typically assumed to be normally distributed: SEM • Error variance is typically assumed to be normal as well • May be applied to binary or ordinal data • Threshold model

  7. What is the best way to measure factors? • Use a sum score • Use a factor score • Use neither - model-fit

  8. FactorScore Estimation • Formulae for continuous case • Thompson 1951 (Regression method) • C = LL’ + V • f = (I+J)-1L’V-1x • Where J = L’V-1 L

  9. FactorScoreEstimation • Formulae for continuous case • Bartlett 1938 • C = LL’ + V • fb = J-1L’V-1x • where J = L’V-1 L • Neither is suitable for ordinal data

  10. Estimate factor score by ML Want ML estimate of this F1 l1 l6 M4 M1 M2 M3 M5 M6

  11. ML FactorScoreEstimation • Marginal approach • L(f&x) = L(f)L(x|f) (1) • L(f) = pdf(f) • L(x|f) = pdf(x*) • x* ~ N(V,Lf) • Maximize (1) with respect to f • Repeat for all subjects in sample • Works for ordinal data too!

  12. Multifactorial Threshold Model Normal distribution of liability x. ‘Yes’ when liability x > t t  0.5 0.4 0.3 0.2 0.1 0 -4 -3 -2 -1 0 1 2 3 4 x

  13. Item Response Theory - Factor model equivalence • Normal Ogive IRT Model • Normal Theory Threshold Factor Model • Takane & DeLeeuw (1987 Psychometrika) • Same fit • Can transform parameters from one to the other

  14. ItemResponseProbability Example item response probability shown in white Response Probability 1 .5 0.5  0.4 1 0.3 .75 0.2 .5 0.1 .25 .0 -4 -3 -2 -1 0 1 2 3 4

  15. Do groups differ on a measure? • Observed • Function of observed categorical variable (sex) • Function of observed continuous variable (age) • Latent • Function of unobserved variable • Usually categorical • Estimate of class membership probability • Has statistical issues with LRT

  16. Practical: Find the Difference(s) 1.00 Variance Mean Mean 1 1 Factor Factor l1 l2 l3 l1 l2 l3 Item 1 Item 2 Item 1 Item 2 r1 r2 Item 3 r3 r1 r2 Item 3 r3 mean1 mean2 mean3 mean1 mean2 mean3 1 1

  17. Sequence of MNI testing 1. Model fx of covariates on factor mean & variance 2. Model fx of covariates on factor loadings & thresholds * If factor loadings equal 3. Revise scale 1 beats 2? Measurement invariance: Sum* or ML scores Yes No MNI: Compute ML factor scores using covariates 2. Identify which loadings & thresholds are non-invariant

  18. Continuous Age as a Moderator in the Factor Model 1.00 b - Factor Variance d - Factor Mean j - Factor Loadings k - Item means v - Item variances [dk and bj confound] d 1 O5 L5 b 0.00 Age 1.00 Factor l3 l2 j l1 1.00 Age V Stims r1 Tranq r2 W5 Z4 MJ r3 mean2 k mean3 mean1 1

  19. What is the best way to measure and model variation in my trait? • Behavioral / Psychological characteristics usually Likert • Might use ipsative? • What if Measurement Invariance does not hold? • How do we judge: • Development • GxE interaction • Sex limitation • Start simple: Finding group differences in mean

  20. Simulation Study (MK) • Generate True factor score f ~ N(0,1) • Generate Item Errors ej ~ N(0,1) • Obtain vector of j item scores sj = L*fj + ej • Repeat N times to obtain sample • Compute sum score • Estimate factor score by ML

  21. Two measures of performance • Reliability

  22. Two measures of performance • Validity

  23. Simulation parameters • 10 binary item scale • Thresholds • [-1.8 -1.35 -0.9 -0.45 0.0 0.45 0.9 1.35 1.8] • Factor Loadings • [.30 .80 .43 .74 .55 .68 .36 .61 .49]

  24. Mess up measurement parameters • Randomly reorder thresholds • Randomly reorder factor loadings • Blend reordered estimates with originals 0% - 100% ‘doses’

  25. Measurement non-invariance • Which works better: ML or Sum score? • Three tests: • SEM - Likelihood ratio test difference in latent factor mean • ML Factor score t-test • Sum score t-test

  26. MNI figures

  27. More Factors: Common Pathway Model

  28. More Factors: Independent Pathway Model = 1 = 1 = 1 Independent pathway model is submodel of 3 factor common pathway model

  29. Example: Fat MZT MZA

  30. Results of fitting twin model

  31. ML A, C, E or P Factor Scores • Compute joint likelihood of data and factor scores • p(FS,Items) = p(Items|FS)*p(FS) • works for non-normal FS distribution • Step 1: Estimate parameters of (CP/IP) (Moderated) Factor Model • Step 2: Maximize likelihood of factor scores for each (family’s) vector of observed scores • Plug in estimates from Step 1

  32. Business end of FS script

  33. The guts of it ! Residuals only

  34. Shell script to FS everyone

  35. 1 Gene  3 Genotypes  3 Phenotypes 2 Genes  9 Genotypes  5 Phenotypes 3 Genes  27 Genotypes  7 Phenotypes 4 Genes  81 Genotypes  9 Phenotypes Central Limit TheoremAdditive effects of many small factors

  36. Measurement artifacts • Few binary items • Most items rarely endorsed (floor effect) • Most items usually endorsed (ceiling effect) • Items more sensitive at some parts of distribution • Non-linear models of item-trait relationship

  37. Assessing the distribution of latent trait • Schmitt et al 2006 MBR method • N-variate binary item data have 2N possible patterns • Normal theory factor model predicts pattern frequencies • E.g., high factor loadings but different thresholds • 0 0 0 0 • 0 0 0 1 but 0 0 1 0 would be uncommon • 0 0 1 1 • 0 1 1 1 • 1 1 1 1 1 2 3 4 item threshold

  38. Latent Trait (Factor) Model Use Gaussian quadrature weights to integrate over factor; then relax constraints on weights F1 Discrimination l1 l6 M4 M1 M2 M3 M5 M6 Difficulty

  39. Latent Trait (Factor) Model Difference in model fit: LRT~ 2 F1 Discrimination l1 l6 M4 M1 M2 M3 M5 M6 Difficulty

  40. Chi-squared test for non-normality performs well

  41. Detecting latent heterogeneityScatterplot of 2 classes S1 Mean S1|c2 Mean S1|c1 S2 Mean S2|c1 Mean S2|c2

  42. Scatterplot of 2 classesCloser means S1 Mean S1|c2 Mean S1|c1 S2 Mean S2|c1 Mean S2|c2

  43. Scatterplot of 2 classesLatent heterogeneity: Factors or classes? S1 S2

  44. Latent Profile Model Class Membership probability Class 1: p Class 2: (1-p)

  45. 1.00 1.00 F F lm lm l1 l1 l2 l2 l3 l3 S1 S1 S2 S2 S3 S3 Sm Sm e1 e1 e2 e2 e3 e3 e4 e4 Factor Mixture Model Class Membership probability Class 1: p Class 2: (1- p) NB means omitted

  46. Classes or Traits? A Simulation Study • Generate data under: • Latent class models • Latent trait models • Factor mixture models • Fit above 3 models to find best-fitting model • Vary number of factors • Vary number of classes • See Lubke & Neale Multiv Behav Res (2007 & In press)

More Related