Re: SEM: Standardized path weight greater than 1?

Could you "draw" the model, or at least tell if the DV is latent or
manifest?

If we are talking about loadings (arrows go from a latent variable to
a manifest one), a loading above 1 is surely erratic - it's like if
you had a correlation coefficient above 1. But, if it's a path
coefficient = regression coefficient (from manifest to latent or
manifest-manifest or latent-latent) its nothing so strange. It just
means that if predictor value grows by 1 its st dev the dependent
value grows by more than one its std dev. Why not? Only check if the
standard error of the parameter is not strange.



On Oct 26, 1:56 am, Richard Ulrich <Rich.Ulr...@xxxxxxxxxxx> wrote:
On Thu, 25 Oct 2007 15:03:32 -0000, "srbraithwa...@xxxxxxxxx"

<srbraithwa...@xxxxxxxxx> wrote:
I have a structural equations model that has one exogenous dichotomous
predictor, two mediators and a dependent variable. The paths from one
of the two indicators that make up the DV loads at 1.06 (the other
loads at 46; these are standardized loadings). What does this mean?

I hope you are talking about a regression coefficient.

A standardized regression coefficient (beta) greater than 1.0
demonstrates that a suppressor relationship exists. It might
be more intelligible and robust, in the long run, to model
that variable and its suppressor as one predictor, looking
at their difference as a precomputed variable.

There's probably something on this in my stats-FAQ.
Or Google-search Groups <group:sci.stat.* suppressor > .

--
Rich Ulrich, wpi...@xxxxxxxxxxxx://www.pitt.edu/~wpilib/index.html


.

Relevant Pages

  • Re: SEM: Standardized path weight greater than 1?
    ... If we are talking about loadings (arrows go from a latent variable to ... a manifest one), a loading above 1 is surely erratic - it's like if ... you had a correlation coefficient above 1. ...
    (sci.stat.edu)
  • Re: basic description of PCA terms
    ... > loadings and scores are all part of the factor analysis framework. ... > "loadings", these are the weights on the latent variables, telling ... Sorry, Data Matter. ... That is THE Principal Component, ...
    (sci.stat.math)
  • Re: Questions about Common Lisp
    ... It means the types are not manifest in the source but are known ... implicitly from and carried by the data ("carried by" meaning data is ... tagged - languages with manifest types may also tag data). ... manifest and latent typing can be analyzed statically. ...
    (comp.lang.scheme)