necessary to determine if these sub-optimally matched subjects may have affected the data in Table 3 (i.e., if those results were spurious). Therefore, Table 4 repeats the analyses from Table 3, but is limited to the 308 optimally ethnically matched individuals. Results in Table 4 are similar to Table 3, showing significant LR × education group × time interactions for both maximum and usual drinks, with between small and medium np2.
