In many code examples, the significance must be lowered for one set of reasons as well as raised for another set of reasons. Typically such adjustments are complex but we are able to estimate them sufficiently to ensure that they result in a net increase in significance beyond what we report on the site. The basic idea follows:

Lowering

Usually, a particular kind of phenomenon is measured, but other kinds might have also been possible. Overall significance therefore decreases if all of these alternatives prove to be insignificant. However, what we often observe instead is the following.

Raising

The unexplored alternatives often turn out to be significant, so that they raise the overall significance rather than lowering it.