AVE & composite reliability-different with different mod

[ruchi]

Hi

I developed a model with 5 latent variables(LV) and one dependent variable (DV)each measured by 3 items
The relation connected in the model 1 is like

LV 1--> DV
LV 2-->DV
LV3-->DV
LV3-->LV1
LV4 -->LV2
LV5-->LV2

Now when i run this using PLS with data metric as mean 0 and Var 1, and then see the default report->quality criterion-->overview, I am able to see composite reliability and AVE.

But when i change my model (all latent variables and dependent variable same as in model 1) as

LV 1--> DV
LV 2-->DV
LV3-->DV
LV3-->LV1
LV4 -->LV1
LV5-->LV2

and then run this model 2 using PLS i am getting different AVE and composite reliability in efault report->quality criterion-->overview.

I think AVE and composite reliability should be same irrespective of the model as it is the validity of the construct.

Kindly help/guide me regarding AVE and composite reliability.

Thanks
Ruchi

[Diogenes]

Hi

some differences are normal, because the “partial” aspect of the algorithm, in one step the outer model (measurement) is estimated, in the next the inner model (structural) and continues until the convergence is reached.

If the structural model change, all values could change.

Best regards,

Bido