can i compare not significant value and significant value?

#1
situation : there are four variables, including exogenous variable x, mediator m1 and m2, and endogenous variable y.

x m1 y
m2

1. I did mediation analysis on two paths
x --> m1 --> y
x --> m2 --> y
(i did sobel test)

the sobel test results indicate that x>m1>y path is significant, but x>m2>y path is not significant.


in this situation,
can i do the test to compare effects of x>m1>y and x>m2>y, when one path is insignificant?
(I am thinking of doing Wald chisquares test)

sorry for my broken english! Thank you!!
 
#2
I am replying to myself..
Basically, significance test means testing if a test value is 0 or not. Therefore, there is no problem comparing nonsignificant coefficient a and significant coefficient b.
 

Dason

Ambassador to the humans
#3
I didn't quite understand your original question but I don't think the conclusion you made:
Therefore, there is no problem comparing nonsignificant coefficient a and significant coefficient b.
follows from what you've presented...
 
#4
Oh I just wondered if I could compare the two path coefficients when one is 'insignificant' (p > 0.05).
But I realized that 'significance test' of a coefficient is testing if the size of coefficient is 0 or not, (Null hypothesis H0: b=0)
Therefore, there's no problem to compare an insignificant coefficient and a coefficient significant regarding which one is larger..


Thank you