Change in predictive power of a model

Questions about the implementation and application of the PLS-SEM method, that are not related to the usage of the SmartPLS software.
Post Reply
rpalrecha
PLS Junior User
Posts: 10
Joined: Sun Nov 19, 2006 4:04 am
Real name and title:

Change in predictive power of a model

Post by rpalrecha »

Hello,

My question is related to change in R2 and how to test the significance of this change in SmartPLS.

My research design is related to testing the predictive power of leadership theories. The data was collected from one organization, and the same respondents have filled out measures for independent and dependents variables of the model (leader behavior and follower performance). The sample size is the same for all three model.

In order to test the predictive power of the model, I did following steps:

1) Run model A, B, and C independently, where model A, B represent two different leadership theories and model C is derived from interview data.

2) Reduce the models based on path coefficients, i.e. retain the variables which are significant or near significant.

3) Run the reduced models A, B, and C independently to make sure that the predictive power of model is consistent with the original model.

4) Run the reduced model A and add variables from model B to look at change in R2. Ran other possible combinations (Model (A+C), added model B etc.)

I have two questions -
a) Am I violating any assumptions by running only the reduced model?
b) How do I check for the significance of change in R2 in SmartPLS?


Thanks
Rita Palrecha
Rita Palrecha
User avatar
Diogenes
PLS Super-Expert
Posts: 899
Joined: Sat Oct 15, 2005 5:13 pm
Real name and title:
Location: São Paulo - BRAZIL
Contact:

Post by Diogenes »

Hi,

a) No, it is OK.

b) For the addition of q independent (exogenous) variables:
df = n - p - 1 and Delta_F = Delta_R2 * (n - p - 1) / (q * (1 - R2_last))

SmartPLS doesn't do this, you must compute by "hand".

see more details in the p.599 of SPSS 15.0 Algorithms available in https://www.washington.edu/uware/spss/docs/

Best regards

Bido
rpalrecha
PLS Junior User
Posts: 10
Joined: Sun Nov 19, 2006 4:04 am
Real name and title:

Predictive power of a model

Post by rpalrecha »

Hello Professor Bido,

Thanks a lot for your reply.

Regards,
Rita Palrecha
Post Reply