Path coefficient and t-value dont match

DrNeurocooper · Post by **DrNeurocooper** » Fri Jul 12, 2013 3:54 pm

Dear PLS community,

I have a following issue with my path coefficients and their significance in my structural model:

Exogenous variable A (formative) has a path coef. of -0,07 and t-value of 2,38;

Exogenous variable B (only one indicator) has a path coef. of 0,23 and t-value of 0,95.

Endogenous variable would be C. The relationships are the following:
A->C
B->C

Can you tell me if I need to drop those two exogenous variables?

Thanks in advance

Hengkov · Post by **Hengkov** » Fri Jul 12, 2013 9:45 pm

Hi,

This is a bug software. I already mention that in old post.
If beta negative, T-statistic also negative, because beta/standar error = T-statistic. So, you T-statistic NOT POSITIVE, but negative. This is potential misinterpret (Type I error).

Regards,

DrNeurocooper · Post by **DrNeurocooper** » Sat Jul 13, 2013 9:40 am

Dear Hengky,

thanks a lot - it was new to me about the bug.

What about the second case, where
"Exogenous variable B (only one indicator) has a path coef. of 0,23 and t-value of 0,95"?
How should I interpret this one? Is it still not significant and this variable has to be droped?

Regards

DrNeurocooper · Post by **DrNeurocooper** » Tue Jul 16, 2013 3:55 pm

Can someone help me with an answer for me?
I would appreciate if someone would give me a reference on this topic?
Thanks in advance.

Hengkov · Post by **Hengkov** » Fri Jul 19, 2013 11:58 pm

Hi,

Why drop, misspecification model? Still retain & not sig. But if you used NHST, cannot say not sig., but not influence.

Regards,

DrNeurocooper · Post by **DrNeurocooper** » Sat Jul 20, 2013 12:25 pm

Thanks a lot.
So the hypothesis will not be droped, but just noted that they have an influence, but not significant.

I need it for my master thesis, so till now only basic knowledge of PLS is there.

Thanks again