“Ghost Chasing”: Demystifying Latent Variables and SEM
description
Transcript of “Ghost Chasing”: Demystifying Latent Variables and SEM
![Page 1: “Ghost Chasing”: Demystifying Latent Variables and SEM](https://reader035.fdocuments.us/reader035/viewer/2022070405/5681401c550346895dab70b1/html5/thumbnails/1.jpg)
1
“Ghost Chasing”:Demystifying Latent Variables and SEM
Andrew Ainsworth
University of California, Los Angeles
![Page 2: “Ghost Chasing”: Demystifying Latent Variables and SEM](https://reader035.fdocuments.us/reader035/viewer/2022070405/5681401c550346895dab70b1/html5/thumbnails/2.jpg)
2/20/2006 Latent Variable Models 2
Topics“Ghost Chasing” and Latent Variables
What is SEM?
SEM elements and Jargon
Example Latent Variables
SEM Limitations
![Page 3: “Ghost Chasing”: Demystifying Latent Variables and SEM](https://reader035.fdocuments.us/reader035/viewer/2022070405/5681401c550346895dab70b1/html5/thumbnails/3.jpg)
2/20/2006 Latent Variable Models 3
“Ghost Chasing”Psychologists are in the business of Chasing “Ghosts”
Measuring “Ghosts”
“Ghost” diagnoses
Exchanging one “Ghost” for another “Ghost”
Latent (AKA “Ghost”) VariablesAnything we can’t measure directly
We must rely on measurable indicators
![Page 4: “Ghost Chasing”: Demystifying Latent Variables and SEM](https://reader035.fdocuments.us/reader035/viewer/2022070405/5681401c550346895dab70b1/html5/thumbnails/4.jpg)
2/20/2006 Latent Variable Models 4
What is a Latent Variable?An operationalization of data as an abstract construct
A data reduction method that uses “regression like” equations
Take many variables and explain them with a one or more “factors”
Correlated variables are grouped together and separated from other variables with low or no correlation
![Page 5: “Ghost Chasing”: Demystifying Latent Variables and SEM](https://reader035.fdocuments.us/reader035/viewer/2022070405/5681401c550346895dab70b1/html5/thumbnails/5.jpg)
2/20/2006 Latent Variable Models 5
Establishing Latent VariablesExploratory Factor Analysis
Summarizing data by grouping correlated variables
Investigating sets of measured variables for underlying constructs
Often done near the onset of research and/or scale construction
![Page 6: “Ghost Chasing”: Demystifying Latent Variables and SEM](https://reader035.fdocuments.us/reader035/viewer/2022070405/5681401c550346895dab70b1/html5/thumbnails/6.jpg)
2/20/2006 Latent Variable Models 6
Establishing Latent VariablesConfirmatory Factor Analysis
Testing whether proposed constructs influence measured variables
When factor structure is known or at least theorized
Often done when relationships among variables are known
![Page 7: “Ghost Chasing”: Demystifying Latent Variables and SEM](https://reader035.fdocuments.us/reader035/viewer/2022070405/5681401c550346895dab70b1/html5/thumbnails/7.jpg)
2/20/2006 Latent Variable Models 7
LatentVariable
Variable1
Variable3
Variable2
Variable4
Conceptualizing Latent Variables
![Page 8: “Ghost Chasing”: Demystifying Latent Variables and SEM](https://reader035.fdocuments.us/reader035/viewer/2022070405/5681401c550346895dab70b1/html5/thumbnails/8.jpg)
2/20/2006 Latent Variable Models 8
Conceptualizing Latent VariablesLatent variables – representation of the variance shared among the variables
common variance without error or specific variance
TotalVariance
CommonVariance
UniqueVariance
SpecificVariance
RandomError
![Page 9: “Ghost Chasing”: Demystifying Latent Variables and SEM](https://reader035.fdocuments.us/reader035/viewer/2022070405/5681401c550346895dab70b1/html5/thumbnails/9.jpg)
2/20/2006 Latent Variable Models 9
What is SEM?SEM – Structural Equation ModelingAlso Known As
CSA – Covariance Structure AnalysisCausal ModelsSimultaneous EquationsPath AnalysisConfirmatory Factor AnalysisLatent Variable Modeling
![Page 10: “Ghost Chasing”: Demystifying Latent Variables and SEM](https://reader035.fdocuments.us/reader035/viewer/2022070405/5681401c550346895dab70b1/html5/thumbnails/10.jpg)
2/20/2006 Latent Variable Models 10
SEM in a nutshellCombination of factor analysis and regression
Tests relationships variables
Specify models that explain data with few parameters
Flexible - Works with continuous and discrete variables
Significance testing and model fit
![Page 11: “Ghost Chasing”: Demystifying Latent Variables and SEM](https://reader035.fdocuments.us/reader035/viewer/2022070405/5681401c550346895dab70b1/html5/thumbnails/11.jpg)
2/20/2006 Latent Variable Models 11
Goals in SEMHypothesize a model that:
Has a number of parameters less than the number of unique Variance/Covariance entries (i.e. (p*(p+1))/2)
Has an implied covariance matrix that is not significantly different from the sample covariance matrix
Allows us to estimate population parameters that make the sample data the most likely
![Page 12: “Ghost Chasing”: Demystifying Latent Variables and SEM](https://reader035.fdocuments.us/reader035/viewer/2022070405/5681401c550346895dab70b1/html5/thumbnails/12.jpg)
2/20/2006 Latent Variable Models 12
Important Matrices
s matrix Sample CovariancesThe data
matrixModel Implied Covariances
Residual Covariance Matrix
Item1 Item2 Item3 Item4
Item1
Item2
Item3
Item4
Item1 Item2 Item3 Item4
Item1 s211 s2
12 s213 s2
14
Item2 s221 s2
22 s223 s2
24
Item3 s231 s2
32 s233 s2
34
Item4 s241 s2
42 s243 s2
44
![Page 13: “Ghost Chasing”: Demystifying Latent Variables and SEM](https://reader035.fdocuments.us/reader035/viewer/2022070405/5681401c550346895dab70b1/html5/thumbnails/13.jpg)
2/20/2006 Latent Variable Models 13
SEM JargonMeasurement
The part of the model that relates measured variables to latent factors
The measurement model is the factor analytic part of SEM
StructureThis is the part of the model that relates variable or factors to one another (prediction)
If no factors are in the model then only path model exists between measured variables
![Page 14: “Ghost Chasing”: Demystifying Latent Variables and SEM](https://reader035.fdocuments.us/reader035/viewer/2022070405/5681401c550346895dab70b1/html5/thumbnails/14.jpg)
2/20/2006 Latent Variable Models 14
SEM JargonModel Specification
Creating a hypothesized model that you think explains the relationships among multiple variables
Converting the model to multiple equations
Model EstimationTechnique used to calculate parameters
E.G. - Ordinary Least Squares (OLS), Maximum Likelihood (ML), etc.
![Page 15: “Ghost Chasing”: Demystifying Latent Variables and SEM](https://reader035.fdocuments.us/reader035/viewer/2022070405/5681401c550346895dab70b1/html5/thumbnails/15.jpg)
2/20/2006 Latent Variable Models 15
SEM JargonModel Identification
Rules for whether a model can be estimated
For example, For a single factor:At least 3 indicators with non-zero loadingsno correlated errorsFix either the Factor Variance or one of the Factor
Loadings to 1
![Page 16: “Ghost Chasing”: Demystifying Latent Variables and SEM](https://reader035.fdocuments.us/reader035/viewer/2022070405/5681401c550346895dab70b1/html5/thumbnails/16.jpg)
2/20/2006 Latent Variable Models 16
SEM JargonModel Evaluation
Testing how well a model fits the data
Just like with other analyses (e.g. ANOVA) we look at squared differencesSEM looks at the squared difference between the s
and matricesWhile weighting the squared difference depending
on the estimation method (e.g. OLS, ML, etc.)
minpick a ( ) ( ) ' ( )Q s W s
![Page 17: “Ghost Chasing”: Demystifying Latent Variables and SEM](https://reader035.fdocuments.us/reader035/viewer/2022070405/5681401c550346895dab70b1/html5/thumbnails/17.jpg)
2/20/2006 Latent Variable Models 17
SEM JargonModel Evaluation
Even with well fitting model you need to test significance of predictorsEach parameter is divided by its SE to get a Z-
score which can be evaluatedSE values are calculated as part of the estimation
procedure
![Page 18: “Ghost Chasing”: Demystifying Latent Variables and SEM](https://reader035.fdocuments.us/reader035/viewer/2022070405/5681401c550346895dab70b1/html5/thumbnails/18.jpg)
2/20/2006 Latent Variable Models 18
Conventional SEM diagrams = measured variable
= latent variable
= regression weight or factor loading
= covariance
![Page 19: “Ghost Chasing”: Demystifying Latent Variables and SEM](https://reader035.fdocuments.us/reader035/viewer/2022070405/5681401c550346895dab70b1/html5/thumbnails/19.jpg)
2/20/2006 Latent Variable Models 19
Sample Variance/Covariance Matrix
X1 X2 X3X1 1.8782 1.0824 1.1080X2 1.0824 2.3414 1.3409X3 1.1080 1.3409 2.6023
![Page 20: “Ghost Chasing”: Demystifying Latent Variables and SEM](https://reader035.fdocuments.us/reader035/viewer/2022070405/5681401c550346895dab70b1/html5/thumbnails/20.jpg)
2/20/2006 Latent Variable Models 20
Basic Tracing Rules for a Latent VariableOnce parameters are estimated
Calculating the Implied Covariance Matrix
Rules for Implied Variance Common Variance – trace a path from a variable back to itself, multiplying parameters
Add to it the unique variance of that DV
Rules for covariance between variablesTrace path from any variable to another, multiplying parameters
![Page 21: “Ghost Chasing”: Demystifying Latent Variables and SEM](https://reader035.fdocuments.us/reader035/viewer/2022070405/5681401c550346895dab70b1/html5/thumbnails/21.jpg)
2/20/2006 Latent Variable Models 21
Implied Covariance Matrix
X1 X3X2
1
2 .9457(1)(.9457) .8944 .9838 1.8782X
2 31.1445(1)(1.1716) 1.3409X X
1 3.9457(1)(1.1716) 1.1080X X
Variances
Covariances
1 2.9457(1)(1.1445) 1.0824X X
2
2 1.1445(1)(1.1445) 1.3099 1.0314 2.3413X
3
2 1.1716(1)(1.1716) 1.3726 1.2296 2.6022X
1.17161.1445.9457
LatentVariable
(1.00)
1.22961.0314.9838
X1 X2 X3X1 1.8782 1.0824 1.1080X2 1.0824 2.3413 1.3409X3 1.1080 1.3409 2.6022
![Page 22: “Ghost Chasing”: Demystifying Latent Variables and SEM](https://reader035.fdocuments.us/reader035/viewer/2022070405/5681401c550346895dab70b1/html5/thumbnails/22.jpg)
2/20/2006 Latent Variable Models 22
Residual Matrix
( )
1.8782 1.0824 1.1080 1.8782 1.0824 1.1080 0 0 0
1.0824 2.3414 1.3409 1.0824 2.3413 1.3409 0 .0001 0
1.1080 1.3409 2.6023 1.1080 1.3409 2.6022 0 0 .0001s residual
![Page 23: “Ghost Chasing”: Demystifying Latent Variables and SEM](https://reader035.fdocuments.us/reader035/viewer/2022070405/5681401c550346895dab70b1/html5/thumbnails/23.jpg)
Function Min and Chi-Square'
'
( )
( ( )) ( ( ))
1.8782 1.0824 1.1080 1.8782 1.0824 1.1080 1 0 0
1.0824 2.3414 1.3409 1.0824 2.3413 1.3409 * 0 1 0
1.1080 1.3409 2.6023 1.1080 1.3409 2.6022 0 0 1
1.8782 1.0824 1.1080
* 1.0824 2.
s W
Q s W s
Q
2
( )
2
1.8782 1.0824 1.1080
3414 1.3409 1.0824 2.3413 1.3409
1.1080 1.3409 2.6023 1.1080 1.3409 2.6022
.00000008
(?) *(492 1) .00000008*491 .00000982
(#unique VC elements)-(# of estimated parame
s
Q
Q
df
2
ters)
6 6 0df
![Page 24: “Ghost Chasing”: Demystifying Latent Variables and SEM](https://reader035.fdocuments.us/reader035/viewer/2022070405/5681401c550346895dab70b1/html5/thumbnails/24.jpg)
2/20/2006 Latent Variable Models 24
Full Measurement Diagram
Depression
BDI
CES-D
ZDRS
NegativeParentalInfluence
Depressparent
InsecureAttach
Neglect
Gender
E
E
E
D
E
E
E
![Page 25: “Ghost Chasing”: Demystifying Latent Variables and SEM](https://reader035.fdocuments.us/reader035/viewer/2022070405/5681401c550346895dab70b1/html5/thumbnails/25.jpg)
2/20/2006 Latent Variable Models 25
SEM limitationsSEM is a confirmatory approach
You need to have established theory about the relationships
Exploratory methods (e.g. model modification) can be used on top of the original theory
SEM is not causal; experimental design = cause
![Page 26: “Ghost Chasing”: Demystifying Latent Variables and SEM](https://reader035.fdocuments.us/reader035/viewer/2022070405/5681401c550346895dab70b1/html5/thumbnails/26.jpg)
2/20/2006 Latent Variable Models 26
SEM limitationsSEM correlational but, can be used with experimental data
Mediation and manipulation can be tested
SEM very fancy technique but it does not make up for a bad methods
![Page 27: “Ghost Chasing”: Demystifying Latent Variables and SEM](https://reader035.fdocuments.us/reader035/viewer/2022070405/5681401c550346895dab70b1/html5/thumbnails/27.jpg)
2/20/2006 Latent Variable Models 27
SEM limitationsBiggest limitation is sample size
It needs to be large to get stable estimates of the covariances/correlations
@ 200 subjects for small to medium sized model
A minimum of 10 subjects per estimated parameter
Also affected by effect size and power
![Page 28: “Ghost Chasing”: Demystifying Latent Variables and SEM](https://reader035.fdocuments.us/reader035/viewer/2022070405/5681401c550346895dab70b1/html5/thumbnails/28.jpg)
2/20/2006 Latent Variable Models 28
Take Home MessagesYou’re a “Ghost Chaser” and didn’t know it
Latent Variables are “Ghosts”
SEM – method for getting closer to studying the “ghosts” directly
SEM is complicated but it is accessible to you if you need to use it
Thank You!!
![Page 29: “Ghost Chasing”: Demystifying Latent Variables and SEM](https://reader035.fdocuments.us/reader035/viewer/2022070405/5681401c550346895dab70b1/html5/thumbnails/29.jpg)
2/20/2006 Latent Variable Models 29
ReferencesBollen, K. (1989) Structural Equations with Latent
Variables. New York : Wiley.
Comrey, A. L., & Lee, H. B. (1992). A First Course in Factor Analysis (2nd ed.). Hillsdale, New Jersey: Lawrence Erlbaum Associates.
Kline, R. B. (1998). Principles and Practice of Structural Equation Modeling. New York: The Guilford Press.
Ullman, J. B. (2001). Structural Equation Modeling. In B. G. Tabachnik & L. S. Fidell (Eds.), Using Multivariate Statistics (4th ed.): Allyn and Bacon.
![Page 30: “Ghost Chasing”: Demystifying Latent Variables and SEM](https://reader035.fdocuments.us/reader035/viewer/2022070405/5681401c550346895dab70b1/html5/thumbnails/30.jpg)
2/20/2006 Latent Variable Models 30
Some SEM advanced questionsAre there group differences?
Multigroup models
e.g. Men vs. Women D
BDI
CES-D
ZDRS
NPI
DP
IA
N
E
E
E
DE
E
E
D
BD
I
CE
S-D
ZD
RS
NP
I
DPIAN
EEE
DEEEMen Women
![Page 31: “Ghost Chasing”: Demystifying Latent Variables and SEM](https://reader035.fdocuments.us/reader035/viewer/2022070405/5681401c550346895dab70b1/html5/thumbnails/31.jpg)
2/20/2006 Latent Variable Models 31
Some SEM advanced questionsCan change in responses be tracked over time?
Latent Growth Curve Analysis
![Page 32: “Ghost Chasing”: Demystifying Latent Variables and SEM](https://reader035.fdocuments.us/reader035/viewer/2022070405/5681401c550346895dab70b1/html5/thumbnails/32.jpg)
2/20/2006 Latent Variable Models 32
Latent Growth Model
QDITime 1
QDITime 2
QDITime 3
QDITime 4
Intercept
1 1 1 1
Slope
0 1 2 3
E1 E2 E3 E4
D1QDI
Time 1QDI
Time 2QDI
Time 3QDI
Time 4D2 D3 D4
Q1 Q2 Q3 Q1 Q2 Q3 Q1 Q2 Q3 Q1 Q2 Q3
E1 E2 E3 E1 E2 E3 E1 E2 E3 E1 E2 E3