R squared and path coefficient

indah kurnia · Post by **indah kurnia** » Sun Jan 08, 2012 6:43 pm

Hi,

I have a problem in interpreting the goodness of fit for structural model in PLS. it is said that the GoF comes from R squared and the significance statistic for path coefficient. I wonder if I have a low R squared 0.071(below 0.1) but the path coefficient is 0.25 and significant in 0.05. Do you think the structural model is still good ?

many thanks,

indah

vonbergh · Post by **vonbergh** » Sun Jan 08, 2012 7:48 pm

Hi Indah,

Based on the assumption that IV -> DV is your model, your R² appears to be small-to-medium (Cohen, 1988) with a significant path coefficient.

Whether your model has a good fit depends on the GoF. A good references to check is

Wetzels, M., Odekerken-Schröder, G., & Van Oppen, C. (2009). Using PLS path modeling for assessing hierarchical construct models: guidelines and empirical illustration. Management Information Systems Quarterly, 33(1), 177-195.

The following threads on GoF may be of further help:
viewtopic.php?t=841
viewtopic.php?t=1094
viewtopic.php?t=1398
viewtopic.php?p=5208#5208

All the best,
Dennis

Mrs Tourky · Post by **Mrs Tourky** » Thu Jan 12, 2012 1:15 pm

Dear Mr Dennis,

Thanks a lot for your contributions on R2 and GoF. They is really helpful and save a lot of time!!

I have a second order formative construct which is formed of two first order formative constructs.

I run algorithm, and R2 for the two first order constructs is Zero; and for the second order constructs is .98!!

I wonder if there is a problem I could address.

Please note, I have checked correlatio (item weight loading). they are all < .7; except one item .69. I removed this item and re-run algorithm, still have got the same zero R2 for first order construct/ and .98 for second order construct.

I really appreciate your time and reply. Thanks

Best,
Marwa