Devi migliorare su questo, ma questo è fondamentalmente il modo in cui lo fai. puoi usare un self-join
SELECT uid, MAX(myvalue)-MIN(myvalue) AS diff FROM
(SELECT a.* FROM tablename AS a
LEFT JOIN tablename AS a2
ON a.uid=a2.uid AND a.myvalue<=a2.myvalue
GROUP BY uid,myvalue
HAVING COUNT(*)<=2
) a
GROUP BY uid