Hi Bob,
Thanks for clarifying. It sounds like you have thought out the conceptual framework quite thoroughly.
Best,
Cam
Cameron N. McIntosh, MA
Analyst / Analyste
Health Analysis and Measurement Group / Groupe d'analyse et de mesure de la santé
Statistics Canada / Statistique Canada
24-Q R.H. Coats Building
Post by Cameron McIntosh100 Tunney's Pasture Driveway
Ottawa, ON
K1A 0T6
Phone: (613) 951-3725
Fax: (613) 951-3959
-----Original Message-----
From: Structural Equation Modeling Discussion Group [mailto:***@BAMA.UA.EDU] On Behalf Of Robert Renn
Sent: September 28, 2006 1:01 PM
To: ***@BAMA.UA.EDU
Subject: Re: Correlated residuals in SEM
Cameron & Holger,
I appreciate both of your helpful comments
on my question about estimating covariances
among residuals in SEM.
I am evaluating a theory I developed around
the question: How does personality relate
to self-management failure? I propose (Renn
et al., 2005, Journal of Management) that
personality works through self-defeating behaviors
to cause self-management failure. I make specific
predictions for how Neuroticism, for example, relates
to procrastination and emotional self-absorption that,
in turn, cause failures in one or all of personal goal
setting, monitoring and operating (how I define self- management in the theory). Thus to test the predictions I prefer to separate the 3 self-management behaviors rather than form a second-order factor.
Because there are causes of self-management excluded from
the causal framework, I thought estimating the covariances among the residuals would be okay. Am I wrong? For instance, I do not include proactive personality or approach or avoidance temperament in the structural equations but they could account for covariance among the self-management behaviors.
I hope this helps somewhat in clarifying what I am doing.
And thank you once again for your guidance.
Bob Renn
Robert W. Renn, Ph.D.
Department of Management
Fogelman College of Business and Economics
The University of Memphis
Memphis, TN 38152
Department Telephone: (901) 678-4551
Direct/Voice Mail: (901) 678-2886
Email: ***@memphis.edu
WWW: https//umdrive.memphis.edu/rrenn/www
________________________________
From: Structural Equation Modeling Discussion Group on behalf of Cameron McIntosh
Sent: Thu 9/28/2006 9:41 AM
To: ***@BAMA.UA.EDU
Subject: Re: Correlated residuals in SEM
Hi Holger,
Yes, that is a excellent point. The higher-order construct in this case is indeed a personality trait rather than a behaviour. However, if we believe that the personality trait is the reason why all three of these behaviours occur together, then we also believe that the covariance matrix of the outcomes is dictated by the variance of the personality trait and the pathweights.... In other words the statistical model suggested seems appropriate for conveying how we think the world works in this case. Cam
Cameron N. McIntosh, MA
Analyst / Analyste
Health Analysis and Measurement Group / Groupe d'analyse et de mesure de la santé Statistics Canada / Statistique Canada 24-Q R.H. Coats Building
Post by Cameron McIntosh100 Tunney's Pasture Driveway
Ottawa, ON
K1A 0T6
Phone: (613) 951-3725
Fax: (613) 951-3959
-----Original Message-----
From: Structural Equation Modeling Discussion Group [mailto:***@BAMA.UA.EDU] On Behalf Of Holger Steinmetz
Sent: September 28, 2006 10:10 AM
To: ***@BAMA.UA.EDU
Subject: AW: Correlated residuals in SEM
Hi Cameron,
I totally agree with your points. But in my view, when a "strong self-managers" is the cause of the behavior, the referring latent is not behavior anymore but the disposistion of this behavior (which can be totally ok).
This is like the distinction between altruistic acts (as compositive behavior construct) and agreeableness as a disposition to perform altruistic behavior.
From the perspective of job performance (where I am currently interested
in) this aspect has been neglected when researchers transformed analysing different performance facets with PCA to CFA. But what is the common variance among different behavior facets?
Best,
Holger
____________________________________________________
Dipl. Psych. Holger Steinmetz
University of Giessen
Department of Work- and Organizational Psychology Otto-Behaghel-Strasse 21 F 35394 Giessen
mail: ***@web.de
DFG-project 'Mobilzeit'
University of Giessen
Institute for Political Science
Department of Empirical Social Research
www.mobilzeit.com ___________________________________________________
-----Ursprüngliche Nachricht-----
Von: owner-***@BAMA.UA.EDU [mailto:owner-***@BAMA.UA.EDU] Im Auftrag von Cameron McIntosh
Gesendet: Mittwoch, 27. September 2006 16:50
An: ***@BAMA.UA.EDU
Betreff: Re: Correlated residuals in SEM
Hi Holger,
Well said, as one definitely needs to carefully consider the direction of influence regarding indicators and constructs that makes the most sense. However, it was my intuition that in this particular case, a strong "self-manager" would display all of the behaviours described. In other words I can see self-management as a unidimensional entity or continuum (high-low) that could conceivably cause one to execute certain behaviours. I guess I asked myself: Does one "become" a good self-manager by just performing a set of requisite behaviours, or is one more or less oriented toward self-management (by whatever is causally upstream from that, personality traits being an obvious one), which is then reflected in these behaviours, across various work and social environments? I suppose both explanations are defensible, but the later struck me as the most plausible. Cam
Cameron N. McIntosh, MA
Analyst / Analyste
Health Analysis and Measurement Group / Groupe d'analyse et de mesure de la santé Statistics Canada / Statistique Canada 24-Q R.H. Coats Building
Post by Cameron McIntosh100 Tunney's Pasture Driveway
Ottawa, ON
K1A 0T6
Phone: (613) 951-3725
Fax: (613) 951-3959
-----Original Message-----
From: Structural Equation Modeling Discussion Group [mailto:***@BAMA.UA.EDU] On Behalf Of Holger Steinmetz
Sent: September 27, 2006 10:24 AM
To: ***@BAMA.UA.EDU
Subject: AW: Correlated residuals in SEM
Hi, Bob and Cameron,
why modeling the outcome as a latent anyway? I have got the opinion that behavior-constructs are almost always formative, unless you have different formulations of the same behavior pattern (e.g., helps colleagues, is altruistic, behaves in a cooperative manner).
Otherwise you are either modeling a) the cognitive schema of the rater regarding the behavior or b) common causes of the behavior (etc. motivation, skills) but no "latent behavior factor".
In their popular article, Law and Wong (1999) discovered such a "trap". They modeled five indicators that described job features (e.g., task significance etc.) as an expression of an "underlying" job description factor. This job description factor was proposed to emerge an effect on job satisfaction. However, a correlation between this factor and job satisfaction of .99 revealed that the job description factor was job satisfaction influencing the five indicators.
In the field of modeling of behavior, common causes are even more likely to be something different than behavior (i.e., any kind of disposition).
Best,
Holger
-------
Law, K. S., & Wong, C.-S. (1999). Multidimensional constructs in structural equation analysis: An illustration using the job perception and job satisfaction constructs. Journal of Management, 25 (2), 143-160.
____________________________________________________
Dipl. Psych. Holger Steinmetz
University of Giessen
Department of Work- and Organizational Psychology Otto-Behaghel-Strasse 21 F 35394 Giessen
mail: ***@web.de
DFG-project 'Mobilzeit'
University of Giessen
Institute for Political Science
Department of Empirical Social Research
www.mobilzeit.com ___________________________________________________
-----Ursprüngliche Nachricht-----
Von: owner-***@BAMA.UA.EDU [mailto:owner-***@BAMA.UA.EDU] Im Auftrag von Cameron McIntosh
Gesendet: Dienstag, 26. September 2006 20:10
An: ***@BAMA.UA.EDU
Betreff: Re: Correlated residuals in SEM
Hi Bob,
Thanks for the clarification. Perhaps a more theoretically meaningful way to model the correlations among the outcomes would be to add a higher-order "self-management" factor, which manifests itself in all 3 of these lower-order constructs? Cam
Cameron N. McIntosh, MA
Analyst / Analyste
Health Analysis and Measurement Group / Groupe d'analyse et de mesure de la santé Statistics Canada / Statistique Canada 24-Q R.H. Coats Building
Post by Cameron McIntosh100 Tunney's Pasture Driveway
Ottawa, ON
K1A 0T6
Phone: (613) 951-3725
Fax: (613) 951-3959
-----Original Message-----
From: Structural Equation Modeling Discussion Group [mailto:***@BAMA.UA.EDU] On Behalf Of Robert Renn
Sent: September 26, 2006 1:57 PM
To: ***@BAMA.UA.EDU
Subject: Re: Correlated residuals in SEM
Cameron,
Thanks for responding to my question.
The outcomes in the model are self-management
behaviors, such as personal goal setting, monitoring
actions against the goals, and operating on self
to achieve the goals.
How does that sound to you in terms of freeing the
parameters estimating the covariances among their residuals?
Bob Renn
Robert W. Renn, Ph.D.
Department of Management
Fogelman College of Business and Economics
The University of Memphis
Memphis, TN 38152
Department Telephone: (901) 678-4551
Direct/Voice Mail: (901) 678-2886
Email: ***@memphis.edu
WWW: https//umdrive.memphis.edu/rrenn/www
________________________________
From: Structural Equation Modeling Discussion Group on behalf of Cameron McIntosh
Sent: Tue 9/26/2006 11:38 AM
To: ***@BAMA.UA.EDU
Subject: Re: Correlated residuals in SEM
Hi Bob,
If you have a good rationale for doing this, go ahead. But first... what exactly are your outcomes? Cam
Cameron N. McIntosh, MA
Analyst / Analyste
Health Analysis and Measurement Group / Groupe d'analyse et de mesure de la santé Statistics Canada / Statistique Canada 24-Q R.H. Coats Building 100 Tunney's Pasture Driveway Ottawa, ON K1A 0T6
Phone: (613) 951-3725
Fax: (613) 951-3959
-----Original Message-----
From: Structural Equation Modeling Discussion Group [mailto:***@BAMA.UA.EDU] On Behalf Of Robert Renn
Sent: September 25, 2006 5:31 PM
To: ***@BAMA.UA.EDU
Subject: Correlated residuals in SEM
SEMnetters,
I am estimating a model with predictors,
mediators and outcomes. For both theoretical
and empirical reasons, the outcomes are
expected to be correlated. Consequently, is it appropriate
to free the parameters representing the residuals of the equations
predicting the outcomes?
I read an example in the LISREL manual that did so, but I
don't have a reference from JAP, AMJ etc for this. If you
do, could you share it with me.
Thank you,
Bob Renn
Robert W. Renn, Ph.D.
Department of Management
Fogelman College of Business and Economics
The University of Memphis
Memphis, TN 38152
Department Telephone: (901) 678-4551
Direct/Voice Mail: (901) 678-2886
Email: ***@memphis.edu
WWW: https//umdrive.memphis.edu/rrenn/www
--------------------------------------------------------------
To unsubscribe from SEMNET, send email to ***@bama.ua.edu with the body of the message as: SIGNOFF SEMNET Search the archives at http://bama.ua.edu/archives/semnet.html
-------------------------------------------------------------- To unsubscribe from SEMNET, send email to ***@bama.ua.eduuu with the body of the message as: SIGNOFF SEMNET Search the archives at http://bama.ua.edu/archives/semnet.html
--------------------------------------------------------------
To unsubscribe from SEMNET, send email to ***@bama.ua.edu
with the body of the message as: SIGNOFF SEMNET
Search the archives at http://bama.ua.edu/archives/semnet.html
--------------------------------------------------------------
To unsubscribe from SEMNET, send email to ***@bama.ua.edu
with the body of the message as: SIGNOFF SEMNET
Search the archives at http://bama.ua.edu/archives/semnet.html
--------------------------------------------------------------
To unsubscribe from SEMNET, send email to ***@bama.ua.edu
with the body of the message as: SIGNOFF SEMNET
Search the archives at http://bama.ua.edu/archives/semnet.html
--------------------------------------------------------------
To unsubscribe from SEMNET, send email to ***@bama.ua.edu
with the body of the message as: SIGNOFF SEMNET
Search the archives at http://bama.ua.edu/archives/semnet.html
--------------------------------------------------------------
To unsubscribe from SEMNET, send email to ***@bama.ua.edu
with the body of the message as: SIGNOFF SEMNET
Search the archives at http://bama.ua.edu/archives/semnet.html
--------------------------------------------------------------
To unsubscribe from SEMNET, send email to ***@bama.ua.edu
with the body of the message as: SIGNOFF SEMNET
Search the archives at http://bama.ua.edu/archives/semnet.html
--------------------------------------------------------------
To unsubscribe from SEMNET, send email to ***@bama.ua.edu
with the body of the message as: SIGNOFF SEMNET
Search the archives at http://bama.ua.edu/archives/semnet.html
--------------------------------------------------------------
To unsubscribe from SEMNET, send email to ***@bama.ua.edu
with the body of the message as: SIGNOFF SEMNET
Search the archives at http://bama.ua.edu/archives/semnet.html