To calculate the values of each cell of a single column, I will have first to tell Power BI to look for the value of the next cell then do a sum or a subtract of the current cell with it by using an index/rank column. I will explain how to do it for a general data then grouped by 1 or multiple conditions. Let´s take an example with this simple data:

First I will create a simple index column for the whole data with this formula and named it that way:

RANKX(ALL('table'),'table'[argument],,ASC) // to sort descending replace ASC by DESC

NOTE:

• Replace “table” and “argument” by yours
• I am using my “number” column to sort ascending instead of a “date” column, “name” column, etc.
• The index can be created in the Power Query Editor (not the purpose here) but when you have a huge data, it won´t work properly

With the index, I will be able to sum each cell with this formula:

var nextrow = 'table'[argument1]+1
var valuenextrow = CALCULATE(VALUES('table'[argument2]),FILTER(ALL('table'),'table'[argument1]=nextrow))
return
CALCULATE(VALUES('table'[argument2])) + valuenextrow // to subtract use this (column 3): valuenextrow - CALCULATE(VALUES('table'[argument2]))

This is the result and the picture on the right, it is just to illustrate my meaning:

NOTE: if you don’t want to use an index and based on your needs, you can get the same result using the OFFSET and/or WINDOW functions (more simple, read this article Power BI: calculate values of a single column by row).

Instead to use a calculated column, I can use a measure. In this case, the formula will be a little different but quite similar:

var nextrow = MIN('table'[argument1])+1
var valuenextrow = CALCULATE(MIN('table'[argument2]),FILTER(ALL('table'),'table'[argument1]=nextrow))
return
CALCULATE(MIN('table'[argument2])) + valuenextrow // to subtract use this (measure 3): valuenextrow - CALCULATE(MIN('table'[argument2]))

Now, let´s say that I want to do that for a particular group (for instance priority). As for the general one, I have to create an index for each priority with this formula:

RANKX(FILTER('table','table'[argument]=EARLIER('table'[argument])),'table'[number],,ASC)

I will start first to calculate for “p5” with this formula:

var nextrow = 'table'[argument1]+1
var valuenextp5 = CALCULATE(VALUES('table'[argument2]),FILTER(ALL('table'),'table'[argument1]=nextrow && 'table'[argument3]="value"))
return
IF([argument3]="value",CALCULATE(VALUES('table'[argument2])) + valuenextp5,BLANK())

If I want to complete the column with the other priorities, just repeat the same formula as “p5” like that:

var nextrow = 'table'[argument1]+1
var valuenextp3 = CALCULATE(VALUES('table'[argument2]),FILTER(ALL('table'),'table'[argument1]=nextrow && 'table'[argument3]="value1"))
var valuenextp4 = CALCULATE(VALUES('table'[argument2]),FILTER(ALL('table'),'table'[argument1]=nextrow && 'table'[argument3]="value2"))
var valuenextp5 = CALCULATE(VALUES('table'[argument2]),FILTER(ALL('table'),'table'[argument1]=nextrow && 'table'[argument3]="value3"))
return
IF([argument3]="value1",CALCULATE(VALUES('table'[argument2])) + valuenextp3,IF([argument3]="value2",CALCULATE(VALUES('table'[argument2])) + valuenextp4,CALCULATE(VALUES('table'[argument2])) + valuenextp5))

I will explain the measure formula for the column 5 because the measure for the column 6 has the same logic:

var nextrow = MIN('table'[argument1])+1
var valuenextp5 = CALCULATE(MIN('table'[argument2]),FILTER(ALL('table'),'table'[argument1]=nextrow && 'table'[argument3]="value"))
return
IF(MIN('table'[argument3])="value",CALCULATE(MIN('table'[argument2])) + valuenextp5,BLANK())

I will complicate a little more, calculating a group of 2 criterias (“group” and “priority”) so again, I will need to create a new index for this group of 2 conditions:

RANKX(FILTER('table','table'[argument1]=EARLIER('table'[argument1]) && 'table'[argument2]=EARLIER('table'[argument2])),'table'[argument3],,ASC)

NOTE: if I need to group more sub-categories, I just need to add another “&& 'table'[argument3]=EARLIER('table'[argument3])”, etc.

The calculation formula (for the column 9 has the same logic):

var nextrow = 'table'[argument1]+1
var valuenextp5ux = CALCULATE(VALUES('table'[argument2]),FILTER(ALL('table'),'table'[argument1]=nextrow && 'table'[argument3]="value1" && 'table'[argument4]="value2"))
var valuenextp5nt = CALCULATE(VALUES('table'[argument2]),FILTER(ALL('table'),'table'[argument1]=nextrow && 'table'[argument3]="value3" && 'table'[argument4]="value2"))
var resultp5ux = CALCULATE(VALUES('table'[argument2])) + valuenextp5ux
var resultp5nt = CALCULATE(VALUES('table'[argument2])) + valuenextp5nt
return
IF([argument3]="value1" && [argument4]="value2",resultp5ux,IF([argument3]="value3" && [argument4]="value2",resultp5nt,BLANK()))

And I will end with the measure (just repeat the same logic for the measure 9):

var nextrow = MIN('table'[argument1])+1
var valuenextp5ux = CALCULATE(MIN('table'[argument2]),FILTER(ALL('table'),'table'[argument1]=nextrow && 'table'[argument3]="value1" && 'table'[argument4]="value2"))
var valuenextp5nt = CALCULATE(MIN('table'[argument2]),FILTER(ALL('table'),'table'[argument1]=nextrow && 'table'[argument3]="value3" && 'table'[argument4]="value2"))
var resultp5ux = CALCULATE(MIN('table'[argument2])) + valuenextp5ux
var resultp5nt = CALCULATE(MIN('table'[argument2])) + valuenextp5nt
return
IF(MIN('table'[argument3])="value1" && MIN('table'[argument4])="value2",resultp5ux,IF(MIN('table'[argument3])="value3" && MIN('table'[argument4])="value2",resultp5nt,BLANK()))