On #4, your confidence interval formula is wrong, inside the tinv function should be 0.05 instead of 0.01. On #5, you are right that the data is matched pair. But the t-stat is wrong because the stdev in the denominator should be calculated from a new variable that captures the difference between 2003 and 2004 for every observation. On #7, the economic meaning of the slope coeff?