The direction of the relationship between an independent and dependent variable was found as positive as measured by Pearson correlation as well as linear regression when no other independent was input. However, when other independents are input into the regression model, the sign of the variable of interest changes to negative. Skimming through some articles, almost all pointed out to multicollinearity as the cause. However, there is no multicollinearity evidenced by (insignificant) weak correlation between the input variables as well as the VIF values (less than 2). What am I missing? How can I explain this?

More Mahyar Masoudi's questions See All
Similar questions and discussions