Abstract
In Path Analysis, especially in social sciences studies, many researchers usually assume that errors in the model are uncorrelated with all exogenous variables as well as with each other. These assumptions, in most cases, are not valid in reality and were introduced to facilitate the model estimation. This article establishes a new algorithm for the computation of the correlation matrix implied by a recursive path model that overcomes these drawbacks. We compare our algorithm to two other methods used in the literature. The comparison was made mathematically through an illustrated example and numerically with a simulation study. The findings show that, unlike the classical methods, the proposed method gives more accurate results.
Similar content being viewed by others
Notes
Bollen (1989) and others choose to define a recursive model by the fact that \({\varvec{B}}\) is lower triangular and \({\varvec{\varPsi }}\) is a diagonal matrix.
more details about the limits of assuming uncorrelated errors are presented in Sect. 2.
Bollen (1989) added the recursiveness of a model is sufficient condition for its identification. However this is based on his definition for recursiveness, which includes a diagonal matrix of \({\varvec{\varPsi }}\).
It should be noted that \({\varvec{0}}\) is used to define the null matrix whose order depends on where it is used. For example: \({\mathbb {E}}[{\varvec{\xi }}{\varvec{\zeta }}^{t}]={\varvec{0}}={\varvec{0}}_{1:q,1:p}\) since \({\varvec{\xi }}\) is vector of order q and \({\varvec{\zeta }}\) is vector of order p. Therefore, The reader shall know from the expression the order of the matrix (eventually vector) \({\varvec{0}}\).
The term augmented set of exogenous variables is used in Kang and Seneta (1980) and it refers to all model variables whose causes are not explicit in the model (i.e. observed exogenous variables and the unmeasured errors.
References
Blalock, H.M.: Correlation and causality: the multivariate case. Soc. Forces 39(3), 246–251 (1961). https://doi.org/10.2307/2573216
Bollen, K.A.: Structural Equations with LV. Wiley, New York (1989)
Brito, C., Pearl, J.: A new identification condition for recursive models with correlated errors. Struct. Equ. Model. Multidiscip. J. 9(4), 459–474 (2002). https://doi.org/10.1207/S15328007SEM0904_1
Duncan, O.D.: Path analysis: sociological examples. Am. J. Soc. 72, 1–16 (1966). https://doi.org/10.1086/224256
Elhadri, Z., Hanafi, M.: The finite iterative method for calculating the correlation matrix implied by a recursive path model. Electron. J. Appl. Stat. Anal. 08(01), 84–99 (2015). https://doi.org/10.1285/i20705948v8n1p84
Elhadri, Z., Hanafi, M.: Extending the finite iterative method for computing the covariance matrix implied by a recursive path model. In: Saporta, G., Russolillo, G., Trinchera, L., Abdi H, Esposito V.V. (eds.),The Multiple Facets of Partial Least Squares Methods: PLS, Paris, France, 2014, Springer, vol. 2, pp. 29–43 (2016)
Elhadri, Z., Mohamed, H., Pasquale, D., Elkettani, Y., Iaousse, M.: Computation of the covariance matrix implied by a structural recursive model with latent variables through the finite iterative method. In: The 5th Edition of the International Conference on Optimization and Applications (ICOA 2019), pp. 272–280 (2019). https://doi.org/10.1109/ICOA.2019.8727648
Greene, W.: Econometric Analysis (Eighth Edition). Pearson (2018)
Griesemer, J.: Must scientific diagrams be eliminable? The case of path analysis. Biol. Philos. 6(2), 155–180 (1991). https://doi.org/10.1007/BF02426836
Iaousse, M., Hmimou, A., El Hadri, Z., El Kettani, Y.: A modified algorithm for the computation of the covariance matrix implied by a structural recursive model with latent variables using the finite iterative method. Stat. Optim. Inf. Comput. 8(2), 359–373 (2020a)
Iaousse, M., Hmimou, A., El Hadri, Z., El Kettani, Y.: On the computation of the correlation matrix implied by a recursive path model. In: 2020 IEEE 6th International Conference on Optimization and Applications (ICOA), pp. 1–5 (2020b)
Jöreskog, K.G.: A general method for the analysis of covariance structures. Biometrica 57, 239–251 (1970). https://doi.org/10.1093/biomet/57.2.239
Jöreskog, K. G.: Structural equation models in the social sciences: Specification, estimation and testing. In: Krishnaiah, R. (ed.) Applications of Statistics, pp. 265–287. North-Holland, Amsterdam (1977)
Kang, K. M., Seneta, E.: Path analysis: an exposition. In: Krishnaiah, P.R. (ed.) Developments in Statistics. Academic Press (1980)
Ke-Hai, Y., Wai, C.: Structural equation modeling with near singular covariance matrices. Comput. Stat. Data Anal. 52(10), 4842–4858 (2008). https://doi.org/10.1016/j.csda.2008.03.030
Kenny, D.A.: Correlation and Causality. Wiley, New York (1979)
Lawley, D.N.: The estimation of factor loadings by the method of maximum likelihood. Proc. R. Soc. Edinb. 60, 64–82 (1940)
Loehlin, J., Beaujean, A.A.: Latent Variable Models: An Introduction to Factor, Path, and Structural Equation Analysis, 5th edn. Routledge, Abingdon (2017)
Mortaza, J., Bentler, P.M.: A modified newton method for constrained estimation in covariance structure analysis. Comput. Stat. Data Anal. 15(2), 133–146 (1993). https://doi.org/10.1016/0167-9473(93)90188-Y
Pugesek, B.: Structural Equation Modeling Applications in Ecological and Evolutionary Biology. Cambridge University Press, New York (2003)
Scheiner, S., Gurevich, J.: Design and Analysis of Ecological Experiments, 2nd edn. Oxford University Press, Oxford (2001)
Shipley, B.: Exploratory path analysis with applications in ecology and evolution. Am. Nat. 149(6), 1113–1138 (1997). https://doi.org/10.1086/286041
Shipley, B.: Cause and Correlation in Biology: A User’s Guide to Path Analysis, Structural Equations, and Causal Inference in R. Cambridge University Press, Cambridge (2016)
Wooldridge, J.: Econometric Analysis of Cross Section and Panel Data, 2nd edn. The MIT Press, Cambridge (2010)
Wright, S.: Correlation and causation. J. Agric. Res. 20(17), 557–585 (1921)
Wright, S.: The theory of path coefficients—a reply to niles’ criticism. Genetics 8(3), 239–255 (1923)
Wright, S.: The method of path coefficients. Ann. Math. Stat. 5(3), 161–215 (1934)
Wright, S.: Path coefficients and path regressions. Alternative or complementarity concepts? Biometrics 16(2), 189–202 (1960). https://doi.org/10.2307/2527551
Yau, L., Lee, S., Poon, W.: Covariance structure analysis with three-level data. Comput. Stat. Data Anal. 15(2), 159–178 (1993). https://doi.org/10.1016/0167-9473(93)90190-5
Acknowledgements
The authors gratefully acknowledge the support of editors during the publication process as well as the useful feedback provided by two anonymous referees.
Author information
Authors and Affiliations
Corresponding author
Ethics declarations
Conflict of interest
The authors declare that they have no conflict of interest.
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Appendices
Appendix 1: Proof of Theorem 2
Consider a recursive path model define by the equation (1) or equivalently (19) and let \({\varvec{\widehat{R}^{\star }}}\) be the implied correlation matrix of the whole model (including disturbances). Let \({\varvec{\widehat{R}^{new\star }}}\) be the matrix computed using Algorithm (2).
We will proceed by an induction to prove that each block \({\varvec{\widehat{R}^{\star }}}_{1:q+p+j,1:q+p+j}\) corresponds to the block \({\varvec{\widehat{R}^{new\star }}}_{1:q+p+j,1:q+p+j}\) for \(j=0,\ldots ,p\).
For \(j=0\), by the definition of the matrices:
We suppose that, for j in \(\{1,\ldots ,p-1\}\) :
Now we will prove it for \(j+1\) (i.e. \({\varvec{\widehat{R}^{new\star }}}_{1:q+p+j+1,1:q+p+j+1}={\varvec{\widehat{R}^{\star }}}_{1:q+p+j+1,1:q+p+j+1}\)).
Since the tow matrices are symmetric and with diagonal elements all equal to 1 (\({\varvec{\widehat{R}^{new\star }}}\) according to Algorithm (2) and \({\varvec{\widehat{R}^{\star }}}\) because it is a correlation matrix), showing that :
is reduced to proving :
According to Algorithm (2):
And by the hypothesis of induction (29):
Hence:
On the other hand, from (20):
And from (19):
Since the matrix \({\varvec{B}}\) is strictly lower triangular: \({\varvec{B}}_{j+1,k}=\beta _{j+1,k}=0\) for \(j+1 \leqslant k \leqslant p\). This also means that \({\varvec{A^{\star }}}_{j+1,q+p+k}=0\) for \(j+1 \leqslant k \leqslant p\) or equivalently :
Hence, from (36):
Using (37), we can write (35) as :
Therefore,
From (34) and (39), we deduce that:
Finally:
Appendix 2: Proof of Corollary 1
Let :
-
\({\varvec{\widehat{R}^{\star }}}\) (respectively \({\varvec{\widehat{R}}}\)) be the correlation matrix of a path model including error terms (respectivlly excluding error terms).
-
\({\varvec{\widehat{R}^{new \star }}}\) (respectively \({\varvec{\widehat{R}^{new}}}\)) be the matrix computed using Algorithm (2)(respectivlly defined by (22)).
According to Theorem (2):
Then from the definition of \({\varvec{\widehat{R}^{new}}}\):
On the other hand, from (19 ):
Hence, from (22):
Therefore:
Rights and permissions
About this article
Cite this article
Iaousse, M., El Hadri, Z., Hmimou, A. et al. An iterative method for the computation of the correlation matrix implied by a recursive path model. Qual Quant 55, 897–915 (2021). https://doi.org/10.1007/s11135-020-01034-1
Published:
Issue Date:
DOI: https://doi.org/10.1007/s11135-020-01034-1