> But on these grounds his argument is shaky. As Clark Glymour puts it,
> "Gould claims that factor analysis produces conjectures about the existence
> of unobserved properties solely because the properties, if they existed,
> would explain features of data; in his phrasing factor analysis 'reifies'
> unobserved quantities, and he thinks 'reification' is a Big Mistake. I
> wonder whether he thinks atoms and molecules and their weights are Big
> Mistakes as well, and if not, why not." (Scientists Respond the Bell Curve,
> p. 259)
>
For once Clark has dropped the ball. Gould has no objection to unobservable properties that explain observed data. His point is that there's no more explanatory reason to draw the vectors towards g than to lots of other points that exqually "explain" the data in the statistical sense, i.e., account for the variance. The reason to doubt the ezxistence of g is that it's arbitrary, not that it's an unobservable. It's hard for me to believe that Clark, of all people, missed this point.
--jks