Negative cross loadings?

[MattiasOtto]

Hi!
I am trying to evaluate a measurement model with Smart PLS. I look at the cross loadings table because I want to know if individual manifest variables really belong in the latent variables I have (somewhat gratuitously ) put then in...
I see that in most cases, it was as I expected the manifests load overwhelmingly on the latents they are supposed to load on.....

...and..

In some cases I see that the loadings on the other latent variables are negative, with a minus sign..in some cases a large number, with a minus sign..

Questions:
a) Is that a good thing or a bad thing?
b) should a cross loading value, be close to zero, and not stray in either direction, positively or negatively?
c) can a cross loading value also be a large negative value, without it meaning that the manifest loads onto that latent?
d) what does it mean if several of the "other" loadings i.e. other than one it "should" be, are negative?

I googled the issue and read something about "..not catching all the opposition"..but I do not understand that phrase..

Please? :-)

Mattias Svahn

[MattiasOtto]

Now I have done some more googling....I seem to understand that in factor analysis a loading of -0.7 is equally relevant as a loading of (+) 0.7...the difference being in how the respondents score...

..is that so also in a cross loadings table in Smart PLS?

?? :-)

[Hengkov]

Hi,

1. For Cross-loading table, just see your variable per indicators (not another construct).
2. Cross-loading value cannot zero or negative but positive only.
3. No
4. Outer loading cannot negative, (I think you have negative question).

Cross-loading for asses discriminant validity, for another method used Fornell-Lacker criterion: SQRT AVE > Correlations.

Best Regards,
Hengky

[MattiasOtto]

Thanx! :-)