This was interesting. I wasn’t sure how PERCENTILEX.INC handled blanks, so I created three test cases from your small sample data set below. Note: I also changed the percentile from 75% to 25%, since 75% is high enough to skip over all the blanks in the data set, and thus with all the non-blank durations equal to one, the filtered and nonfiltered results are the same, which is not a generalizable outcome - just a quirk of the particular sample data set provided. Dropping the figure down to 25% illustrates the difference in how the function handles the three following cases:
Case #1 - sample table is unfiltered, and blanks are left as blanks
Case #2 - sample table is filtered to exclude all rows where [Duration] is blank
Case #3 - sample table where blanks are replaced by zeros, unfiltered
If for purposes of your analysis, blanks are not considered values, then Case #2 is the correct result.
Here’s the measure I used for Case #2:
25 Percentile Duration Filtered =
VAR NoBlanksTable =
FILTER(
SELECTCOLUMNS('Schedule Analytics Blanks',
"Project", 'Schedule Analytics Blanks'[Project Name],
"Duration Elapsed", 'Schedule Analytics Blanks'[Duration]
),
[Duration Elapsed]<>BLANK()
)
RETURN
PERCENTILEX.INC (
NoBlanksTable,
[Duration Elapsed],
0.25
)
Full solution file attached below.
I hope this is helpful.
- Brian
eDNA Forum - Percentile.pbix (43.6 KB)