I have a variable x (a subscale of my IV scale i.e., sexual abuse by mothers) and three variables i.e., y1, y2, y3 & y4 (DV and its three sub scales i.e, behavioral problems, Internalizing problems, externalizing problems and other problems) these variables have near zero correlation with IV i.e., -.04 -.03, -.04, & -.04 that is insignificant.But when this IV is added in a step for hierarchical regression, it transformes into a suppresstant variable and starts predicting the DVs.What should be done; should this IV be excluded from the model on the basis of poor correlation? or it should be kept as a part of regression and later be explained as suppressant variable?