Get nth value from a dataset

Ok, I see what you mean...

How about:

EDIT: Instead of creating the Rank for car in step 1 below, it will probably work using a much simpler:

1) Index for Rank =SUM(Table[car]) * 10000 + SUM(Table[Index])

which avoids needing two RANKX- you will need to take into account the number of rows for the *10000 = or make it really huge like * 1000000000

1) Create a measure combining car rank and index as follows:

Index for Rank =
VAR _t =
    RANKX ( ALLSELECTED ( 'Table' ), [Sum car],, ASC, DENSE )
VAR _V =
    _t * 1000
        + SUM ( 'Table'[index] )
RETURN
    _V

Now the rank measure to use to choose the nth value:

Rank = 
RANKX(ALLSELECTED('Table'), [Index for Rank], ,ASC,Dense)

New file attached

Correct!

but use the method

Index for Rank =SUM(Table[car]) * 100000000000 + SUM(Table[Index])

as the first step (to avoid using the costly RANKX), and then use RANKX over this [Index for Rank].

In theory, using the car value * 10000000000 "provides" the correct order; adding the index makes each value unique.