I run a panel data model (N=5, T=17, unbalance). The depended variable is stationary while three of the six independent variables are non-stationary. These non-stationary variables are cointegrated. Are the regression results misleading?