Reverse coding!

quaddusm · Post by **quaddusm** » Tue Feb 26, 2008 1:20 am

Dear All,

I am new to SMART PLS. I created a new model and ran it successfully with SMART PLS. But after I removed some low loading items of a construct (as par regular literature on PLS) one path loading from this construct to the final dependent construct changed sign (became positive from original negative). It does not make any sense to me!! All I did was removed four low loading items of a construct. Everything else remained the same. The results of the revised model is also same as before except the sing of this structural path loading. In PLS GRAPH Prof Chin calls it reverse coding, which I am a bit confused with! Can anybody shed some light on this? Thanks.

==Mohammed

Diogenes · Post by **Diogenes** » Thu Feb 28, 2008 11:20 am

Hi Mohammed,
How many items remain in the LV?
All items are in the same scale?
You try to run a factor analysis of these items (just them) to see if you have more than one factor?

Usually we said that in the reflective LV mode we could drop items with lower loadings, but if we drop many items, we could ending with items that reflect another LV.

Best regards.
Bido

quaddusm · Post by **quaddusm** » Wed Mar 05, 2008 4:44 am

Dear Bido,

The original LV had 9 items. All items are in the same scale. After running PLS I deleted four low loading items. This resuletd in reversing the weights and loading of the remaining five items and the path loading from this LV to the dependent LV changed sign also (became 0.339 fromoriginal -0.330).

Regards

==Mohammed

Diogenes · Post by **Diogenes** » Thu Mar 06, 2008 12:51 pm

Hi Mohammed,

It looks that the 4 delected items were measuring another LV..!?
Try to run a principal component analysis with only these nine items and see how they are grouped.

Best regards.
Bido