Dave Kent, thanks for answering my question about the "null to zero"
version of addition.

Re the particular example, plate appearances, it occurred to me that we
may have all the data necessary to calculate rather than estimate PA,
except for interference and obstruction. That is, the null values for PA
components AB(none), BB(none), HBP, SH, SF may represent not missing data
but league-seasons in which the scoring category did not exist.

In MS Access lahman52, I looked at the records where HBP is null. I found
the subset for seasons before the hit batsman rule, ten 2004 records, and
one 1916 record --the first of three 1916 stints for Earl Hamilton,
hamilea01
(That stint is not listed on the baseball-reference webpage.)

I don't know whether these stray nulls have been corrected in the databank.

The general question is, have we thoroughly checked the incidence of nulls
in the databank?

Paul Wendt