🚀 go-pugleaf

In article <1993Jul5.192725.1@orion.yorku.ca>, as000047@orion.yorku.ca (Phillip Phan) writes:
> 
> 	On an unrelated topic. Can someone give me a quick explanation why
> 	one would do path analysis using OLS regression rather than go to
> 	a package like LISREL for causal modelling?  I've been told that
> 	LISREL requires a large number of cases.  If I did a factor analysis
> 	of my scales and then used the resulting factors for the path analysis
> 	aren't I getting the same thing as an output from LISREL?
	
   1. One normally works with samples with Ns of 100-200+.  Some work has
been done on the "small sample" properties of LV (e.g., Lisrel) models, but
in general smaller samples are not recommended.
  2. Re: factor analysis, scale construction, using resulting factors in 
path analysis.  No, you would not obtain the same results, with the non-
LISREL approach:
     a) there is a certain indeterminacy in the construction of factor 
scores ;
      b) the f.a. + path analysis approach does not correct for 
attenuation that occurs as a result of random measurement error (in a system
of equations with more than 2 variables, the "bias" is not always in the
direction of reducing the magnitude of the estimated parameters in relation
to their "real" values (i.e., that which would have been obtained had 
measurement error not been present).
      c) the Lisrel approach allows one to adjust for certain forms of 
*non-random* measurement error (or test for presence).
---------------------------------------------------------------------
Douglas Baer, Department of Sociology,
University of Western Ontario, London, Canada N6A 5C2
Internet: BAER@UWOVAX.UWO.CA