Statistics • 263
a reasonable level. A test p value that exactly reaches the signicance
level of p = .05 will reach the correct result nearly 85% of the time; this is
resolution.
We can see a fundamental tension in hypothesis testing. e more resis-
tant a test is to Type I errors, the less sensitive it is; in other words, the more
likely it is we will miss real relationships too weak to cross the threshold
set by the signicance level. A real relationship may hit p = .006, and if our
signicance level is set at p = .005, we will reject that real level. However, if
we set a looser threshold as our signicance level, such as p = .05, then we
will churn up more trivial results that will clog our analytical resources
at best and lead us to propose ...