Since the data in this report came from a sample, a sampling error
(standard error) is associated with each reported number. In general,
if the difference between 2 numbers is greater than twice the standard
error for that difference, there is confidence that for 95
out of 100 possible samples a real difference exists and that the
apparent difference is not simply the result of using a sample rather
than the entire population. All differences discussed in the text of
this report were statistically significant at or above the 95-percent
confidence level. Standard error estimates were generated by using a
bootstrap method (jackknife) available for WESVAR PC.