How to explain significant correlation between independent and dependent variable but non significant regression test in the presence of mediation? 
Possible Duplicate:
Not-significant F but a significant coefficient in multiple linear regression 

In my research, I use Optimism as the IV, Job Satisfaction as the DV and Work-Family Enrichment as Mediator.
Path Model Analysis using AMOS shows that IV and DV are significantly correlated, but regression test was non-significant.
What conclusion can we draw from this?
 A: Jerzy Neyman found a positive correlation between number of storks sighted (IV) and number of babies born (DV), he created a dataset (not real data) to illustrate your situation where if you only look at storks and babies you see a relationship, but if you adjust for another variable (size of county measured by number of women in the county) the relationship goes away.  This is because the size and characteristics of the county are most likely the cause of both storks (they like to nest in areas that people like to live in) and babies, but storks and babies are independent once you adjust for the county size.
This is a common theme in statistics, where 2 things appear related (and can often humorously be questioned about cause and effect), but both are really the result of a third cause.  Some common examples include the number of methodist ministers and the sale of rum were highly correlated in one dataset (this was during a time of population growth in the area which likely caused both), or number of fire fighters and the cost of repairs to the building (does this mean we should send fewer fire fighters?).  
