Blank Is Not NULL: A Port Data-Quality Report

Before modelling, a data team profiles missing values by embarkation port. Two kinds of missing are hiding in passengers: age is NULL for 177 passengers, and one passenger's embarked is an empty string — not NULL, just ''.

Show exactly these 5 columns, in this order:

- port — embarked, except that a NULL or blank port becomes 'Unknown'
- passengers — everyone from that port
- ages_missing — passengers from that port with no age
- avg_age_known — the average of the ages that are recorded, rounded to 1 decimal
- avg_age_filled — the average after each missing age is filled with the average age of passengers in the same pclass and the same sex, computed over the whole ship rather than per port, rounded to 1 decimal

Order by passengers descending, then port.

Three NULL rules meet in this report. COALESCE only replaces NULL — it leaves '' exactly where it was, so a blank port has to be turned into NULL first. AVG ignores NULLs, which is why avg_age_known needs no filter and why an average over each pclass and sex group is a clean fill value. And COUNT(age) counts known ages, not passengers.

The fill has to be computed before you group by port. Group first and the class-and-sex averages are gone.