Max value with a subcategory

Anonymous 

you can create a calculated column

column=

var _max=CALCULATE(max('Table'[begin date]),ALLEXCEPT('Table','Table'[ag_level]))

return if('Table'[begin date]=maxx(FILTER('Table','Table'[ag_level]=EARLIER('Table'[ag_level])&&'Table'[begin date]<_max),'Table'[begin date]),1)

 

 

filter the column to 1, then no matter you choose daily or weekly , it will only show the data that you want.

Hi Anonymous 

Try the following measures:

All Second Highest Dates = 
VAR _perPartition =
    SUMX (
        VALUES ( 'Table'[ag_level] ),
        VAR _secondDate =
            MAXX (
                INDEX (
                    2,
                    ALL ( 'Table'[ag_level], 'Table'[begin date] ),
                    ORDERBY ( [begin date], DESC ),
                    ,
                    PARTITIONBY ( 'Table'[ag_level] )
                ),
                [begin date]
            )
        RETURN
            SUMX (
                FILTER ( 'Table', 'Table'[begin date] = _secondDate ),
                'Table'[enrolled_count]
            )
    )
RETURN
    _perPartition

Daily Second Highest Dates = 
VAR _perPartition =
    SUMX (
        VALUES ( 'Table'[ag_level] ),
        VAR _secondDate =
            MAXX (
                INDEX (
                    2,
                    ALL ( 'Table'[ag_level], 'Table'[begin date] ),
                    ORDERBY ( [begin date], DESC ),
                    ,
                    PARTITIONBY ( 'Table'[ag_level] )
                ),
                [begin date]
            )
        RETURN
            SUMX (
                FILTER ( 'Table', 'Table'[begin date] = _secondDate && 'Table'[ag_level] = "Daily" ),
                'Table'[enrolled_count]
            )
    )
RETURN
    _perPartition

Weekly Second Highest Dates = 
VAR _perPartition =
    SUMX (
        VALUES ( 'Table'[ag_level] ),
        VAR _secondDate =
            MAXX (
                INDEX (
                    2,
                    ALL ( 'Table'[ag_level], 'Table'[begin date] ),
                    ORDERBY ( [begin date], DESC ),
                    ,
                    PARTITIONBY ( 'Table'[ag_level] )
                ),
                [begin date]
            )
        RETURN
            SUMX (
                FILTER ( 'Table', 'Table'[begin date] = _secondDate && 'Table'[ag_level] = "Weekly" ),
                'Table'[enrolled_count]
            )
    )
RETURN
    _perPartition

Details are in the attached pbix