Extracting ranked values from different rows - SQL - sql-server

Question

I have a database with various categories. For each category I have three quantities, and I want to extract a row containing the 25th largest value from each of the quantities per category (ties can be safely ignored).
For example, I might have a database whose rows were towns or cities from one of several countries. The categories are countries, and the quantities might be population, land area, and latitude. The data would then look something like:
TownName Country Population LandArea Latitude
Paris France 500,715 47.9 45.76
Manchester USA 110,229 90.6 42.99
Calais France 72,589 33.5 50.95
Leicester England 337,653 73.3 52.63
Dunkirk France 90,995 43.9 51.04
... ... ... ... ...
In this example, the end result I'd want would be each of the countries in the list, along with their 25th largest population, 25th largest land area and 25th largest latitude. This no longer resembles some specific town or city, but gives some information about each country. This might look like:
Country Population LandArea Latitude
France 144,548 83.95 50.21
Poland 141,080 88.3 54.17
Australia 68,572 146 -21.35
... ... ... ...
I've figured out one way to do this, which was to do the following:
Use the ROW_NUMBER function to rank one of Population, LandArea and Latitude in descending order, partitioned over countries.
Repeat this three times (one for each quantity), and JOIN the three databases together. In the ON statement, ensure the values of the Country columns are equal, as are the values of the rank columns.
Use a WHERE statement to pull out the row for each country with rank 25.
I don't like this method because it involved creating three almost exact copies of decent sized chunks of code to get three separate databases I joined together (each of the blocks of code in the join statments were a decent size because this is a simplified example, and I had to do other stuff to get to a stage like this).
I was wondering whether there was a way which wouldn't involve me repeating large chunks of code with a JOIN statement as this makes my code big and ugly. Also, this seems like something which may crop up time and time again, so a more efficient method would be wonderful.
Thanks for your time

Answer 1

Perhaps if you can't find a way to eliminate the 3-join approach, you can simplify the join condition by assigning each distinct tuple a GroupID:
;WITH
MasterCTE AS
(
SELECT *,
DENSE_RANK() OVER (ORDER BY Country) AS GroupID -- Don't use ROW_NUMBER here. RANK or DEMSE_RANK only
FROM MyTable
),
cte1 AS
(
SELECT GroupID, [Population],
ROW_NUMBER() OVER (PARTITION BY GroupID ORDER BY [Population] DESC) AS PopulationRank
FROM MasterCTE
),
cte2 AS
(
SELECT GroupID, LandArea,
ROW_NUMBER() OVER (PARTITION BY GroupID ORDER BY LandArea DESC) AS LandAreaRank
FROM MasterCTE
),
cte3 AS
(
SELECT GroupID, Latitude,
ROW_NUMBER() OVER (PARTITION BY GroupID ORDER BY Latitude DESC) AS LatitudeRank
FROM MasterCTE
)
SELECT DISTINCT -- Remember to include DISTINCT
MasterCTE.Country,
cte1.Population, cte2.LandArea, cte3.Latitude
FROM MasterCTE
INNER JOIN cte1 ON MasterCTE.GroupID = cte1.GroupID AND cte1.PopulationRank = 25
INNER JOIN cte2 ON MasterCTE.GroupID = cte2.GroupID AND cte2.LandAreaRank = 25
INNER JOIN cte3 ON MasterCTE.GroupID = cte3.GroupID AND cte3.LatitudeRank = 25