Whether calculated (age at an event), estimated (e.g. average age at marriage, implied as "before" or "after" a date by activity such as buying land or needing a guardian), or selected from conflicting options, it should be documented.
The dates we pick take on a life of their own, and can have unintended consequences, contributing to the creation of duplicated profiles, encouraging conflation, and even altering research avenues of subsequent researchers.
Some "calculations" can be complex involving interpretation beyond simple math. A person's census records over several decades may result in several different calculated birth years, sometimes differing by a decade or more.
The other answers posted provide several alternatives, each good. I would add that the complexity of the "calculation" should be considered. The census example might be best discussed in a Vital Statistics research note where each listed age can be discussed, as the discussion reflects OUR research and analysis of the sources, rather than the person's actual life.
A simple case like a person stating their age in their Will or in a court record works well directly in the Biography, and personalizes an often stale recounting of records.
I think use of the Estimated Date Research Note Box (defined as a "rough estimate") should always be accompanied by a research note. And if we replace an existing date, the prior date should be documented in the research notes to save subsequent researchers from having to repeat our analysis, and discourage destructive back and forth as one date is replaced with another date for usually mysterious reasons.