Moreover a regression is inappropriate for this sort of analysis. A Structural Equation Modelling approach or a path analysis approach is more appropriate. One thing they did use was a recursive SEM - very iffy since it makes many of the endogenous variables, ie., predictors with no incoming paths, related. Therefore the influence of any direct paths from the endogenous to the dependent variables are problematic because of the uncontrolled correlation between at least 2 of the endogenous variables. Another problem is that they are basing their analysis on an OLS approach, and then morphing it into their specialized spacial analysis. In my opinion an OLS appropriate is not appropriate for their data, even when corrected for widely differing variance amoung counties (heteroskedasticity).
One thing the authors did not do is report the correlation matrix - on the beta coefficients and their associated t-values. So there is no way to easily replicate their results. Moreover they also did not report on the theta values or the errors - again not good - its hard to judge how adequate their analysis is, their measures may be so filled with reliability errors that their results. finally some of their mapping plots show more relationship with population size and location and less to do with any influence by Walmart. In other words, while this article does show some support, it is so weak that I would discount it. larry On 1/19/06, Nick McClure <[EMAIL PROTECTED]> wrote: > One of the problems with that study IIRC they only counted counties that had > greater than 1500 workers in the county. So they had already skewed their > results by not including small counties. > > > > _____ > > From: dana tierney [mailto:[EMAIL PROTECTED] > Sent: Wednesday, January 18, 2006 9:14 PM > To: CF-Community > Subject: Re: [signs of sanity] MD no longer subsidizing Wal-Mart > > > > If the University of PA study I posted late last night doesn't do this to > your satisfaction, then I don't know what to say. I do think it would take > regression analysis and though I have limped through one or two a while back > it would take more time and resources than I personally have available to do > a complete model. I'll check the other thread to see what if anything you > have to say about it. > > I have to admit though, I am getting tired of the topic and I suspect > everyone else is too. > > Dana > > > > > > ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~| Message: http://www.houseoffusion.com/lists.cfm/link=i:5:193155 Archives: http://www.houseoffusion.com/cf_lists/threads.cfm/5 Subscription: http://www.houseoffusion.com/lists.cfm/link=s:5 Unsubscribe: http://www.houseoffusion.com/cf_lists/unsubscribe.cfm?user=11502.10531.5 Donations & Support: http://www.houseoffusion.com/tiny.cfm/54
