Chih-Mao Hsieh schrieb: > If what is meant by linearity is basically left truncation, then I have understood > it... > Hmmm, I don't want to be the spokesman of your pre-poster. But I use this as "linearity of composition of the variables", such that you assume, that the resulting measures are *linear* combinations of the underlying factors. That excludes data, where the influence of the factors is quadratic, cubic etc.
So you could have [1] measure = a*factor1 + b*factor2 + ... but not [2] measure = a*factor1^2 + b*factor2^3 + c*factor1*factor2 + ... In one of the historical example of Thurstone's Ball-Problem, geometrical measures of squared and cubed influences (length and width determining the measures surface. volume, weight) were involved, and the recovering of the factors happened not to be too convincing, if I understand some articles correctly. The basic assumtion of factor analysis is equation [1], thus factor analysis is basically a "linear" model, and I expect someone talking about "linearity" in the context of factor analysis is meaning just this. Gottfried Helms . . ================================================================= Instructions for joining and leaving this list, remarks about the problem of INAPPROPRIATE MESSAGES, and archives are available at: . http://jse.stat.ncsu.edu/ . =================================================================
