scalar max of measure/dynamically filtering insignificant sources

[Anonymous]

Hi folks. 

I am trying to clean up some graphs by filtering out sources with insignificant contribution.  Example, with y-values a measure:

the "blank" sources are just small. Because the amounts vary substantially by selection/navigation, I can't filter the graphic by a static number like 100 or even 10k because depending on source selection, it would still show insignificant sources or no data at all.

If I could create a measure filter_measure = log10([max of shown measure]/[shown measure]). then I can filter the graphic to only show sources within several orders of magnitude of the max value in the graph. Eg: [filter_measure] < 3.

I can't figure out how to calculate the "[max of shown measure]" part, though. I can't use calculated columns because the max is selection-specific.  I've tried a lot of the solutions using maxx, but all of them evaluate to 1 because maxx is evaluated within the context of the selection, so MAXX[shown measure] = [shown measure]. I've seen some solutions with calculate(), but I'm not adding a filter. I thought about using summarize because I'm trying to convince PowerBI to "forget" the row-based context for the purposes of calculating this max and generate a table with only a single row corresponding to whatever source is generating the max.

I figure there's gotta be an easier way. 

[Anonymous · Accepted solution]

A clever peer gave me the answer: MAXX(ALL([table])[shown measure] made everything magically work. Here's my final filtering measure = 
if([shown_measure] = 0, 10, LOG10(MAXX(ALL[table]),[shown measure])/[shown measure])

and graphic filter set final filtering measure < 3.  

[Syk]

Try something like this

max measure = VAR SummarizedTable =
SUMMARIZE(
    'Table',
    'Table'[Category],
    "Sum of Values", SUM('Table'[Value])
)
return
maxx(SummarizedTable,[Sum of Values])