I realize this post may be stale, but if you do not need flexibility with the filter, you do not need to duplicate the data source.
An example... say we have a stacked bar split by some True/False dimension IsMember. Then we want a dual axis line with compound growth (or whatever calculation you want). We want to show calculation for a subset, say where IsMember = TRUE.
We can just hard code it into the calculated field references before aggregations.
BEFORE (shows compound growth for entire stacked bar)
POWER(ZN(SUM([Number of Records]))/LOOKUP(ZN(SUM([Number of Records])), FIRST()),ZN(1/(INDEX()-1))) - 1
AFTER (filtering by hypothetical dimension IsMember)
POWER(ZN(SUM(IF([IsMember]) = TRUE THEN [Number of Records] END))/LOOKUP(ZN(SUM(IF([IsMember]) = TRUE THEN [Number of Records] END)), FIRST()),ZN(1/(INDEX()-1))) - 1
in the same graph, can I add another filter to change both the values?
This is an old (2013-14) and tortured (table calculations) way of doing this. There's probably a better way using LOD expressions. I suggest you start a new thread, and post a detailed explanation and workbook for what you're trying to accomplish.