Path analysis

From InterSciWiki
Jump to: navigation, search


Contents

Mar 2nd Doug: R SUR software for Wikipedia:Seemingly unrelated regressions and path analysis

http://cran.r-project.org/web/packages/systemfit/index.html 

URL:	http://www.systemfit.org
NeedsCompilation:	no

Citation: systemfit citation info - Arne Henningsen and Jeff D. Hamann. 2007. systemfit: A Package for Estimating Systems of Simultaneous Equations in R. Journal of Statistical Software 23(4), 1-40. URL http://www.jstatsoft.org/v23/i04/.

A key feature of a statistically valid NoVA model is that DEf and other regression models eliminate what might be spurious variables that are not predictive given the independent variables that contribute to total variance and that pass tests of exogeneity and other criteria. A NoVA graph also makes all variables endogenous because any of the error terms of constitutent DEf models for dependent variables may be correlated with the independent variables of other models. There are, however, corrections for these endogeneities that allow statistical inferences to be made about direct and indirect effects Wikipedia:Seemingly unrelated regressions for which the SUR R script is available (Henningsen and Hamann 2007).


Corresponding BibTeX entry: 

 @Article{,
   title = {systemfit: A Package for Estimating Systems of
     Simultaneous Equations in R},
   author = {Arne Henningsen and Jeff D. Hamann},
   journal = {Journal of Statistical Software},
   year = {2007},
   volume = {23},
   number = {4},
   pages = {1--40},
   url = {http://www.jstatsoft.org/v23/i04/},
 }
In views:	Econometrics, Psychometrics, SocialSciences
CRAN checks:	systemfit results

Feb 5th 2013 Malcolm Reply to Doug the 3rd Feb

Doug: good to hear that you found someone to help with the path analysis/sem modeling. I look forward to seeing what Ken and his student come up with.

I’m not sure what issues you have that you want to discuss with me, but I’ll be happy to hear them and respond.

The Garson book looks pretty good. My own experience with path analysis is based on a somewhat older literature, but it is basically the same with some language tweeks. I’ll take a look at the book soon, though, to bring myself a bit more current. I’ll also take a close look at your new intro for your chapter 5 and see if I have any suggestions.

Malcolm

Feb 3rd 2013 DRW commentary after day's discussion with statistician Ken Koput

Path analysis may be viewed as an extension of multiple regression models in which all variables become endogenous because paths interconnect (in a directed asymmetric graph). If variables in each component regression model are exogenous, meaning uncorrelated with their error term, their error terms may be correlated. In these "seemingly (un)related regressions", formulas for restoring exogeneity Wikipedia:Seemingly unrelated regressions are required. These adjustments may be solved by simultaneous equations Sewall Wright (1921 1923 1934). Without simultaneous equation models (SEM), "path analysis can be accomplished as a series of multiple linear regressions, one for each endogenous variable. This method yielded standardized regression coefficients (beta weights) and a R-squared goodness-of-fit for each endogeneous variable, but did not yield an overall goodness-of-fit for the model", as does SEM. While "SEM typically centers on latent variables, it is possible to model simple observed variables. When only observed variables are in the model, the researchers is conducting a path analysis" (Garson 2012:5). Results will still be highly tentative because R-squared goodness-of-fit is a relative concept that cannot establish that a model is "correct" (2012:5). SEM has the capability of comparing covariance matrices of alternative models for best fit. Path analysis does not, in itself, have this capability. MCMCregress, an R program, does have this capability for each constituent regression model of a path analysis. With imputation of missing data, MCMCregress requires that the regression model (DEf for example), output imputed data. Given this capability, goodness-of-fit for alternative constituent regression models in an observed variables path analysis while adjustment for the effects of endogenous variables between models can improve interpretation of constituent models before attempting SEM. It would seem that this stepwise approach would be advantageous in contributing to falsification of theory that is suitable for testing with simple observed variables, and thus to cross-cultural research with access to a rich inventory of simple observed variables. Doug (talk) 12:39, 3 February 2013 (PST)

References and suggestions provided by Ken Koput

1*http://en.wikipedia.org/wiki/Seemingly_unrelated_regressions -- good for formulas

Garson, David. 2012. Path

Nelson C. Mark, Masao Ogaki, Donggyu Sul. 2003. Dynamic Seemingly Unrelated Congregating Regression Models. Technical Report Series,

Nelson C. Mark, Masao Ogaki, Donggyu Sul. 2003. Seemingly Unrelated Regression models pdf http://ideas.repec.org/p/nbr/nberte/0292.html

2*Hsiao, Cheng. 2003. Analysis of Panel Data.


http://davidakenny.net

http://davidakenny.net/cm/pathanal.htm

Brant and Williams. 2007. CHAPTER 2. SPECIFICATION OF SIMULTANEOUS EQUATION MODELS. In, Multiple Time Series Models. Sage

http://www.nber.org/chapters/c11699.pdf

Paxton, John Hipp, Marquart. 2001. Non-Recursive Models. Sage

Virendera K. Srivastava, David E.A. Giles. Specification of Unrelated Regression Equation Models. (Statistics: A Series of Textbooks and Monographs)

http://davidakenny.net/cm/pathanal.htm

independent paths (mediation)

Older Introduction to SEM

Books on Path Analysis

Path Analysis music-canal-perform
http://www.statisticalassociates.com/pathanalysis.htm

SEM and path analysis software

1 Wirtschaftuniversita ̈t Wien 2 University of California, Los Angeles. Department of Statistics Papers

UCLA Department of Statistics Papers

Department of Statistics Papers

  • Pearl, Judea: Comments and Controversies: Graphical models, potential outcomes and causal inference: Comment on Lindquist and Sobel, 2011
  • Pearl, Judea: Foreword, 2011
  • Pearl, Judea: An Introduction to Causal Inference, 2011
  • Pearl, Judea: The Mediation Formula: A guide to the assessment of causal pathways in nonlinear models, 2011
  • Pearl, Judea: Principal Stratification — a Goal or a Tool?, 2011
  • Pearl, Judea: Statistics and Causality: Separated to Reunite——Commentary on Bryan Dowd’s ‘‘Separated at Birth’’, 2011
  • Pearl, Judea: Transportability across studies: A formal approach, 2011

Books in UC Libraries

Yes there are books with title Path Analysis 1. see http://melvyl.worldcat.org/oclc/1504955 Path analysis : a primer Author: Ching Chun Li Publisher: Pacific Grove, Calif. : Boxwood Press, [1975] It's on the shelf at 4 UC's: Riverside, Santa Barbara, Davis, and Berkeley. Also a copy here in SD county at San Diego State University if you have a courtesy UCSD library card that allows you to use San Diego Circuit and you want to pick up the book here at UCSD Geisel Library.

2. This book has a chapter on path analysis http://roger.ucsd.edu:80/record=b2054491~S9 or http://melvyl.worldcat.org/oclc/216723 Introduction to multivariate analysis for the social sciences Author: Johannes Petrus van de Geer Publisher: San Francisco : W.H. Freeman, 1971. 

Books, Geisel Floor1 East 	 QA278 .G44

Contents: Part 1. Introduction to matrix algebra: -- An overview of matrix algebra -- Basic concepts of matrix algebra -- Geometric representation of matrices -- Determinants and matrix inversion -- Equations -- Matrix differentiation -- Eigenvectors and Eigenvalues -- The multinormal distribution -- Part 2. Techniques of multivariate analysis: -- An overview of multivariate techniques -- Simple regression and correlation -- Multiple and partial correlation -- Partial-correlation analysis and path analysis -- Factor analysis -- Canonical correlation analysis -- Varieties of factor analysis -- Nonrecursive linear models -- Factors and structural equations -- Discriminant analysis -- Appendix: An iterative solution of Eigenvalues and Eigenvectors.

3. This book covering some aspects of path analysis is at UCSD today and doesn't require ILL. Title Latent variable models : an introduction to factor, path, and structural equation analysis / John C. Loehlin Author Loehlin, John C Published Mahwah, N.J. : L. Erlbaum Associates, 2004 Edition 4th ed Location Call number Status 

Books, Geisel Floor1 East 	 QA278.6 .L64 2004

There's 2 dissertations that start with the phrase Path analysis since I am not sure what you're trying to locate:

- A path analysis of the causal elements in Bandura's theory of self-efficacy and an anxiety-based model of avoidance behavior by Deborah Louise Feltz

- Path analysis models of psychosocial adjustment among Southeast Asian immigrant youth by May Lim

If you want books covering the topic or a statistics encyclopedia article/entry, try these 2 e-book sites/works (and use the phrase causal analysis in addition to path analysis):

- Int'l Encyclopedia of Statistical Science at http://link.springer.com/referencework/10.1007/978-3-642-04898-2/page/1

- StatsNetBase at http://www.crcnetbase.com/page/statistics_ebooks

Thank you for using UCSD Ask A Librarian. Let me know if you need more help.

Deborah Kegel

Retrieved from "http://intersci.ss.uci.edu/wiki/index.php?title=Path_analysis&oldid=76194"
Personal tools