path analysis
DESCRIPTION
PATH ANALYSIS. (see Chapters 5,6 of Neale & Cardon). -. allows us to represent. linear. models for the relationships between. variables in diagrammatic form. e.g. a genetic model;. a factor model;. a regression model. -. - PowerPoint PPT PresentationTRANSCRIPT
![Page 1: PATH ANALYSIS](https://reader031.vdocument.in/reader031/viewer/2022012916/56812f2f550346895d94c254/html5/thumbnails/1.jpg)
PATH ANALYSIS
(see Chapters 5,6 of Neale & Cardon)
- allows us to represent linear models for the relationships betweenvariables in diagrammatic form
e.g. a genetic model; a factor model; a regression model.
- makes it easy to derive expectations for the variances and covariancesof variables in terms of the parameters of the proposed linear model.
- permits easy translation into a matrix formulation as used by programssuch as MX.
![Page 2: PATH ANALYSIS](https://reader031.vdocument.in/reader031/viewer/2022012916/56812f2f550346895d94c254/html5/thumbnails/2.jpg)
dcae
P
E A C D
PHENOTYPE
P = eE + aA + cC + dD
e.g. UNIQUEENVIRONMENT
ADDITIVEGENETICEFFECT
SHAREDENVIRONMENT
DOMINANCEGENETICEFFECT
1 1 11
![Page 3: PATH ANALYSIS](https://reader031.vdocument.in/reader031/viewer/2022012916/56812f2f550346895d94c254/html5/thumbnails/3.jpg)
Figure 1. A PATH MODEL FOR MZ TWIN PAIRS
EXOGENOUSVARIABLES
ENDOGENOUSVARIABLES
1.0 (MZT) or 0.0 (MZA)
ET1 AT1 CT1 DT1
dcae dcae
1
PT2PT1
1
1 1 1 1 1 1ET2 AT2 CT2 DT2
1.0
1.0
TWIN 1 TWIN 2
![Page 4: PATH ANALYSIS](https://reader031.vdocument.in/reader031/viewer/2022012916/56812f2f550346895d94c254/html5/thumbnails/4.jpg)
CONVENTIONS OF PATH ANALYSIS
Squares or rectangles denote observed variables.
Circles or ellipses denote latent (unmeasured) variables.
Upper-case letters are used to denote variables.
Lower-case letters (or numeric values) are used to denote covariances or path coefficients.
Single-headed arrows or paths ( ) are used to represent hypothesized causal relationships between variables under a particular model -- where the variable at the tail of the arrow is hypothesized to have a direct causal influence on the variable at the head of the arrow (e.g., AT1 PT1).
![Page 5: PATH ANALYSIS](https://reader031.vdocument.in/reader031/viewer/2022012916/56812f2f550346895d94c254/html5/thumbnails/5.jpg)
Double-headed arrows ( ) are used to represent a covariance between two variables, which may arise through common causes not represented in the model. Double-headed arrows ( ) may also be used to represent the variance of a variable.
Double-headed arrows may not be used for any variable which has one or more single-headed arrows pointing to it -- these variables are called endogenous variables. Other variables are exogenous variables.
Single-headed arrows may be drawn from exogenous to endogenous variables (A P) or from endogenous variables to other endogenous variables(P1 P2).
![Page 6: PATH ANALYSIS](https://reader031.vdocument.in/reader031/viewer/2022012916/56812f2f550346895d94c254/html5/thumbnails/6.jpg)
Omission of a two-headed arrow between two exogenous
variables implies the assumption that the covariance of those
variables is zero (e.g., no genotype-environment correlation).
Omission of a direct path from an exogenous (or endogenous)
variable to an endogenous variable implies that there is no direct
causal effect of the former on the latter variable.
![Page 7: PATH ANALYSIS](https://reader031.vdocument.in/reader031/viewer/2022012916/56812f2f550346895d94c254/html5/thumbnails/7.jpg)
ASSUMPTIONS OF OUR SIMPLE PATH MODEL FOR MZ TWIN PAIRS:
(i) linear additive effects -- no multiplicative (genotype x environment) interactions;
(ii) E, A, C, D are mutually uncorrelated -- no genotype-environment covariance;
(iii) path coefficients e, a, c, d are the same for first, second twins;
(iv) no reciprocal sibling environmental influences; i.e., no direct paths P1 P2
![Page 8: PATH ANALYSIS](https://reader031.vdocument.in/reader031/viewer/2022012916/56812f2f550346895d94c254/html5/thumbnails/8.jpg)
TRACING RULES OF PATH ANALYSIS
(Simplified rules assuming no ‘feedback loops’, i.e., no reciprocal causation -reciprocal sibling environmental influences - etc.)
(1) Trace backwards, change direction at a 2-headed arrow, then traceforwards.
implies that we can never trace through two 2-headed arrows inthe same chain.
(2) The expected covariance between two variables, or the expectedvariance of a variable, is computed by multiplying together all thecoefficients in a chain, and then summing over all possible chains.
![Page 9: PATH ANALYSIS](https://reader031.vdocument.in/reader031/viewer/2022012916/56812f2f550346895d94c254/html5/thumbnails/9.jpg)
CONTRIBUTION
e.g., for the MZ twin pair covariance we have:
a 1 aa2
c 1 c
d 1 d
CovMZ = a2 + c2 + d2
P1 AT1 AT2 P2
P1 CT1 CT2 P2 c2
P1 DT1 DT2 P2 d2
![Page 10: PATH ANALYSIS](https://reader031.vdocument.in/reader031/viewer/2022012916/56812f2f550346895d94c254/html5/thumbnails/10.jpg)
e.g., for the total phenotypic variance for MZ twins:
CONTRIBUTIONe 1 e
e2
a 1 a
c 1 c
d 1 d
VarMZ = e2 + a2 + c2 + d2
P1 ET1 ET1 P1
P1 AT1 AT1 P1 a2
P1 CT1 CT1 P1 c2
P1 DT1 DT1 P1 d2
![Page 11: PATH ANALYSIS](https://reader031.vdocument.in/reader031/viewer/2022012916/56812f2f550346895d94c254/html5/thumbnails/11.jpg)
PATH ANALYSIS: EXERCISES
Exercise 1. Using Figure 1, what is the expected covariance between MZtwin pairs reared apart (MZAs)?
Exercise 2. Using Figure 1, what is the expected variance of twins fromMZ pairs reared apart?
![Page 12: PATH ANALYSIS](https://reader031.vdocument.in/reader031/viewer/2022012916/56812f2f550346895d94c254/html5/thumbnails/12.jpg)
Figure 2. DZ TWIN PAIRS/FULL SIBLINGS
1 (DZT/FST) or 0 (DZA/FSA)
D E A C E A C D
ched dche
1 1 1 1 1 1 1 1
0.25
0.5 (1+x)
x = 0 under random mating.
PTWIN 2PTWIN 1
![Page 13: PATH ANALYSIS](https://reader031.vdocument.in/reader031/viewer/2022012916/56812f2f550346895d94c254/html5/thumbnails/13.jpg)
Exercise 3. Using Figure 2, and assuming random mating (x=0), what isthe expected covariance between DZ twin pairs reared together?
Exercise 4. Using Figure 2, what is the expected variance of DZ twins orfull siblings reared together?
Exercise 5. Using Figure 2, what is the expected covariance of DZ twinsor full sibling pairs reared apart?
![Page 14: PATH ANALYSIS](https://reader031.vdocument.in/reader031/viewer/2022012916/56812f2f550346895d94c254/html5/thumbnails/14.jpg)
Figure 3. BIOLOGICAL PARENT - OFFSPRING
E C A D E C A D
dacedace
0.25
PMOTHER PFATHER
0.50.5
0.5
0.5
E C A D D A C E
dhce echd
PCHILD 2PCHILD 1
RA
1
RA
1
0.5
0.5
1 1 1 1 1 1 1 1
1 1 11 11
1
![Page 15: PATH ANALYSIS](https://reader031.vdocument.in/reader031/viewer/2022012916/56812f2f550346895d94c254/html5/thumbnails/15.jpg)
Exercise 6. Using Figure 3, what is the expected covariance of parent andoffspring?
Exercise 7. Using Figure 3, what is the expected covariance of biologicalparent and adopted-away offspring?
Exercise 8. What 10 assumptions are implied by Figure 3?
![Page 16: PATH ANALYSIS](https://reader031.vdocument.in/reader031/viewer/2022012916/56812f2f550346895d94c254/html5/thumbnails/16.jpg)
Figure 4. BIOLOGICAL UNCLE / AUNT - NEPHEW / NIECE
1
0.5
E C A D E C A D
d
PAUNT/UNCLE
PMOTHER
ace dace
0.25
E C A D
0.5
0.5
PNEPHEW/NIECE
E C A D
dace
RA 0.5
PFATHER
dace
1
![Page 17: PATH ANALYSIS](https://reader031.vdocument.in/reader031/viewer/2022012916/56812f2f550346895d94c254/html5/thumbnails/17.jpg)
Exercise 9. Using Figure 4, what is the expected covariance betweenbiological uncle and nephew?
![Page 18: PATH ANALYSIS](https://reader031.vdocument.in/reader031/viewer/2022012916/56812f2f550346895d94c254/html5/thumbnails/18.jpg)
Figure 5. OFFSPRING OF MZ TWIN PAIRS
E D C A A C D E
d ace
PSPOUSE1
½
E D A C E D A C
cade
RA
0.5
1 1 1 1 1 1 1 1
1 1
1
1
11
PTWIN1 PSPOUSE2
POFFSPRING (Twin 1)POFFSPRING (Twin 2)
E D C A
½
1 1 1 1
c eda
PTWIN2
A C D E1 1 1 1
cade
A
0.5
R
11
1
d ace c eda
½ ½
![Page 19: PATH ANALYSIS](https://reader031.vdocument.in/reader031/viewer/2022012916/56812f2f550346895d94c254/html5/thumbnails/19.jpg)
Exercise 10. Using Figure 5, what is the expected covariance betweenMZ uncle/aunt and nephew/niece, i.e. between an MZ twin and the child of his or her cotwin?
Exercise 11. What is the expected covariance between first cousins related through MZ twins (“MZ half sibs”)?
![Page 20: PATH ANALYSIS](https://reader031.vdocument.in/reader031/viewer/2022012916/56812f2f550346895d94c254/html5/thumbnails/20.jpg)
Figure 6. DIRECTION-OF-CAUSATION MODEL: MZ PAIRS
E C1 1
CONDUCTPROBLEMS TWIN 1
e c
E C1 1
CONDUCTPROBLEMS TWIN 2
e c
1
E A1 1
ALCOHOLDEPENDENCE TWIN 1
e a
E A1 1
ALCOHOLDEPENDENCE TWIN 2
e a
1 ii
![Page 21: PATH ANALYSIS](https://reader031.vdocument.in/reader031/viewer/2022012916/56812f2f550346895d94c254/html5/thumbnails/21.jpg)
Exercise 12. Using Figure 6,
(1) What is the expected MZ covariance for(a) conduct problems;(b) alcohol dependence;(c) conduct problems in one twin and alcohol
dependence inthe cotwin?
(2) Redraw Figure 6 so that the direction of causation is now
ALCOHOL DEPENDENCE CONDUCT
PROBLEMS
What are the expected MZ covariances now for(a) conduct problems;(b) alcohol dependence;(c) conduct problems in one twin and alcohol
dependence inthe cotwin?
![Page 22: PATH ANALYSIS](https://reader031.vdocument.in/reader031/viewer/2022012916/56812f2f550346895d94c254/html5/thumbnails/22.jpg)
(3) What are the corresponding expected DZ covariances when
(a) CONDUCT PROBLEMS ALCOHOL DEPENDENCE
(b) ALCOHOL DEPENDENCE CONDUCT PROBLEMS
![Page 23: PATH ANALYSIS](https://reader031.vdocument.in/reader031/viewer/2022012916/56812f2f550346895d94c254/html5/thumbnails/23.jpg)
PART TWO:
PATH ANALYSIS: FROM PATH DIAGRAMS TO MATRICES
It is straightforward to convert from a path diagram to a matrix representation of a path model, and vice versa. Here we illustrate a computationally efficient approach -though other alternatives exist (see e.g. the MX manual).
Consider first a model with no feedback loops --for example Figure 1 for MZ twin pairs reared together, where all endogenous variables are directly observable, and there are no paths from endogenous variables to other endogenous variables.
Let X denote the covariance matrix for the exogenous variables. The i, j-th element of the matrix X (where i denotes Row, and j denotes Column) will be equal to the covariance of the i-th, j-th exogenous variables. X will be a square matrix.
![Page 24: PATH ANALYSIS](https://reader031.vdocument.in/reader031/viewer/2022012916/56812f2f550346895d94c254/html5/thumbnails/24.jpg)
XMZ =
1.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
0.0 1.0 0.0 0.0 0.0 1.0 0.0 0.0
0.0 0.0 1.0 0.0 0.0 0.0 1.0 0.0
0.0 0.0 0.0 1.0 0.0 0.0 0.0 1.0
0.0 0.0 0.0 0.0 1.0 0.0 0.0 0.0
0.0 1.0 0.0 0.0 0.0 1.0 0.0 0.0
0.0 0.0 1.0 0.0 0.0 0.0 1.0 0.0
0.0 0.0 0.0 1.0 0.0 0.0 0.0 1.0
![Page 25: PATH ANALYSIS](https://reader031.vdocument.in/reader031/viewer/2022012916/56812f2f550346895d94c254/html5/thumbnails/25.jpg)
Let W denote the matrix of paths from the exogenous variables to the endogenous variables. The i, j-th element of the matrix W will be equal to the path from the j-th exogenous variable to the i-th endogenous variable. W will be a rectangular matrix, with number of rows equal to the number of endogenous variables, number ofcolumns equal to the number of exogenous variables.
W = e a c d 0 0 0 00 0 0 0 e a c d
![Page 26: PATH ANALYSIS](https://reader031.vdocument.in/reader031/viewer/2022012916/56812f2f550346895d94c254/html5/thumbnails/26.jpg)
For this simplified case, with no paths from endogenous variables to other endogenous variables, and all endogenous variables directly observable, we can derive the expected covariance matrix as a product of matrices,
MZ = WXMZ W
![Page 27: PATH ANALYSIS](https://reader031.vdocument.in/reader031/viewer/2022012916/56812f2f550346895d94c254/html5/thumbnails/27.jpg)
WX = e a c d 0 a c d
0 a c d e a c d
e a c d 0 a c d
0 a c d e a c dWXW = e 0
a 0
c 0
d 0
0 e
0 a
0 c
0 d
MZ = e2 + a2 + c2 + d2 a2 + c2 + d2
a2 + c2 + d2 e2 + a2 + c2 + d2
![Page 28: PATH ANALYSIS](https://reader031.vdocument.in/reader031/viewer/2022012916/56812f2f550346895d94c254/html5/thumbnails/28.jpg)
Exercise 13. Using Figure 2, write down the elements of the covariance matrix of the exogenous variables for DZ pairs, XDZ.
Derive DZ = WXDZ W
![Page 29: PATH ANALYSIS](https://reader031.vdocument.in/reader031/viewer/2022012916/56812f2f550346895d94c254/html5/thumbnails/29.jpg)
Figure 7 represents a more complicated model, with paths
from endogenous variables to other endogenous variables --
in this example, representing the reciprocal environmental
influences of the twins upon each other (or perhaps a rater
contrast effect, if twin data are ratings by the same parent).
![Page 30: PATH ANALYSIS](https://reader031.vdocument.in/reader031/viewer/2022012916/56812f2f550346895d94c254/html5/thumbnails/30.jpg)
Figure 7. MODEL WITH RECIPROCAL SIBLING EFFECTS
1 (MZT/DZT/FST) or 0 (MZA/DZA/FSA)
D E A C E A C D
caed dcae
1 (MZ) or 0.25 (DZ/FS)
1 (MZ) or 0.5 (DZ/FS)
PTWIN 2PTWIN 1
i - reciprocal sibling interaction effect
0 (MZA/DZA/FSA) or i (MZT/DZT/FST)
0 (MZA/DZA/FSA) or i (MZT/DZT/FST)
![Page 31: PATH ANALYSIS](https://reader031.vdocument.in/reader031/viewer/2022012916/56812f2f550346895d94c254/html5/thumbnails/31.jpg)
Y = 0 i
i 0
Let Y denote the matrix of paths from endogenous variables to other endogenous variables. Yi,j is the path from the j-th variable to the i-th variable, with Yi,i = 0 by definition.
![Page 32: PATH ANALYSIS](https://reader031.vdocument.in/reader031/viewer/2022012916/56812f2f550346895d94c254/html5/thumbnails/32.jpg)
It may be shown (see Neale and Cardon, C6), again assuming that all endogenous variables are directly observable, that
where Z = (I - Y)-1
and
Here I is the identity matrix, and -1 denotes the inverse.
MZ = ZWXMZ W Z
DZ = ZWXDZ W Z
![Page 33: PATH ANALYSIS](https://reader031.vdocument.in/reader031/viewer/2022012916/56812f2f550346895d94c254/html5/thumbnails/33.jpg)
Exercise 14. What are the expected covariance matrices for MZ and DZ twin pairs reared together, and for MZ and DZ twin pairs reared apart, under the model of Figure 7?
![Page 34: PATH ANALYSIS](https://reader031.vdocument.in/reader031/viewer/2022012916/56812f2f550346895d94c254/html5/thumbnails/34.jpg)
Exercise 15. Redraw Figure 6 for the case of reciprocal causation, i.e. a path i from ALCOHOL DEPENDENCE to CONDUCT PROBLEMS, in addition to the path i from CONDUCT PROBLEMS to ALCOHOL DEPENDENCE. What is the expected 4x4 covariance matrix for MZ pairs for this case?
![Page 35: PATH ANALYSIS](https://reader031.vdocument.in/reader031/viewer/2022012916/56812f2f550346895d94c254/html5/thumbnails/35.jpg)
THE END