High t-stat for intercept? I have built a multivariate distributed lag model (two regressors) which produces p-values that are significant just below the one percent level. However, my constant has a t-stat of like 26. Should I have concerns?
The R-squared is like .80.
 A: The constant term, $\beta_0$, or $y$-intercept, is important since when used it means you are not forcing the $y$-intercept to go through zero, i.e., that is, when $x=0 \rightarrow y=0$.  When employed, you are allowing the $y$-intercept to estimate the grand mean of $y$ when $x=0$, and then the slope coefficient $\beta_1$ reveals the change in $y$ for a 1-unit change in $x$.  When you get more advanced with multiple linear regression models for which there are multiple slopes $(\beta_1, \beta_2, \ldots, \beta_p)$, it works out that the constant term will represent the mean value of $y$ for objects (records) for which multiple $x_j=0$.  At this point, most people want the significance of $t_j$ and magnitude (how large each $\beta_j$ is) and the sign of each $\beta_j$ ($j \neq 0$).  Then again, sometimes people drill down into the constant term $\beta_0$ for a multiple linear regression model to reveal what is going on.  However, you don't have to really focus on the constant term if you don't want to.  
If your slope term $\beta_1$ is not significant, then the $x$-variable when lagged does not explain the variation in $y$.
