Selecting time based on the max value of different column

Question

I am doing traffic analysis and I have an aggregated view of average travel times with traffic. What I am attempting to do is select the time associated with the max average travel time for the given time periods.

Here is the code and results I am looking for, without time:

SELECT dt.datekey, dt.peak_hours, dt.weekday_name, 
ROUND(MAX(avg_t_dur_traffic)/60,2) MaxATT
FROM cat.Analytics_AvgTimes att
INNER JOIN cat.Dim_Date dt on att.DateKey = dt.DateKey and att.timekey = dt.timekey
WHERE dt.DateKey = '20170523' AND dt.peak_hours <> 'off_peak'
GROUP BY dt.DateKey, dt.Peak_Hours, dt.Weekday_Name

datekey     peak_hours    weekday_name  MaxATT
2017-05-23  AM_Peak       Tuesday       28.93
2017-05-23  Midday_Peak   Tuesday       14.05
2017-05-23  PM_Peak       Tuesday       29.95

Edit: I'm not looking specifically for these three MaxATT times, I'm only looking for help on how to query the information...

I see the Max Average Trip Time (MaxATT) during AM peak hours on Tuesday approx. 29 minutes. When I add in time is where I am having a hard time. I get the following results:

Please note that I have removed a few rows from the results to save on space

SELECT dt.datekey, dt.TimeKey, dt.peak_hours, dt.weekday_name, 
ROUND(MAX(avg_t_dur_traffic)/60,2) MaxATT
FROM cat.Analytics_AvgTimes att
INNER JOIN cat.Dim_Date dt on att.DateKey = dt.DateKey and att.timekey = dt.timekey
WHERE dt.DateKey = '20170523' and dt.peak_hours <> 'off_peak'
GROUP BY dt.DateKey, dt.Peak_Hours, dt.Weekday_Name, dt.TimeKey

datekey     TimeKey     peak_hours  weekday_name    MaxATT
2017-05-23  05:15:00    AM_Peak     Tuesday         10.43
2017-05-23  05:30:00    AM_Peak     Tuesday         10.45
2017-05-23  07:15:00    AM_Peak     Tuesday         12.53
2017-05-23  07:30:00    AM_Peak     Tuesday         18.27
2017-05-23  07:45:00    AM_Peak     Tuesday         22.85
2017-05-23  08:00:00    AM_Peak     Tuesday         28.93

We can see the time associated with MaxATT = 28.93 is 08:00 during the AM peak. But all I want to select is each distinct time during each peak, so what I am looking for is

datekey     TimeKey     peak_hours    weekday_name  MaxATT
2017-05-23  08:00:00    AM_Peak       Tuesday       28.93
2017-05-23  13:45:00    Midday_Peak   Tuesday       14.05
2017-05-23  17:30:00    PM_Peak       Tuesday       29.95

Here is an attempt at using RANK(), but I failed miserably as MaxATT does not match the correct value, and it pulls more records than just RANK() = 1

SELECT id, datekey, peak_hours, weekday_name, MaxAtt
FROM(
        SELECT id, dt.DateKey, dt.Peak_Hours, dt.Weekday_Name, 
        ROUND(MAX(avg_t_dur_traffic)/60,2) MaxAtt,
        RANK() OVER(PARTITION BY id ORDER BY MAX(avg_t_dur_traffic) DESC) AS 
        [rank]
        FROM cat.Analytics_AvgTimes att
        INNER JOIN cat.Dim_Date dt ON att.datekey = dt.DateKey AND att.TimeKey = dt.TimeKey
        GROUP BY id, dt.DateKey, dt.Peak_Hours, dt.Weekday_Name
     ) mt
WHERE mt.[rank] = 1 AND DateKey = '20170523' AND peak_hours <> 'off_peak'
GROUP BY id, datekey, peak_hours, weekday_name, MaxAtt

Thank you for the help and if you need further clarification please feel free to ask!

Edit: Sample data as requested

Answer 1

Use a subquery to find the maxatt rows needed and they join that to the original table.

select a.* 
from cat.Analytics_AvgTimes a
Inner JOIN (
          select datekey, peak_hours, weekday_name, max(maxatt) as maxatt
          from cat.Analytics_AvgTimes
          group by datekey, peak_hours, weekday_name) b
ON a.datekey = b.datekey and a.peak_hours = b.peak_hours and 
   a.weekday_name = b.weekday_name and a.maxatt = b.maxatt

fiddle example http://sqlfiddle.com/#!9/dcacb9/1/0

edit after reviewing data:

select one.*
from (select a.*, b.Avg_T_Dur_Traffic as max
      from dim_date a, Analytics_AvgTimes b
      where a.datekeyid = b.id) one
INNER JOIN (
      select a.*, max(b.Avg_T_Dur_Traffic) as max
      from dim_date a, Analytics_AvgTimes b
      where a.datekeyid = b.id
      group by datekey, monthid, weekid, dayid, peakhoursid) two
      ON one.datekey = two.datekey and one.monthid = two.monthid and
         one.weekid = two.weekid and one.dayid = two.dayid and one.max = two.max

fiddle with provided data: http://sqlfiddle.com/#!9/d7889a/9