Creating a correlation matrix for two multi-value variables?

You're on the right track, sort of.  First step is to create all permutations between x and y values. The index column is interesting but not really needed (except maybe for the self correlation you mentioned).

So you go from

to 

by using two simple splits

let
    Source = Table.FromRows(Json.Document(Binary.Decompress(Binary.FromText("NU7LFcAwCNrFM4dG82lnycv+azQQc9AnCOicVgxWELs7AtUWpjnRrsCAiwmhLlUXU+lTEbU9kalouae2wdVP6mBG3jmu9zjkJf5ScfflyUcUun4=", BinaryEncoding.Base64), Compression.Deflate)), let _t = ((type nullable text) meta [Serialized.Text = true]) in type table [Index = _t, #"Variable X" = _t, #"Variable Y" = _t]),
    #"Split Column by Delimiter" = Table.ExpandListColumn(Table.TransformColumns(Source, {{"Variable X", Splitter.SplitTextByDelimiter(",", QuoteStyle.Csv), let itemType = (type nullable text) meta [Serialized.Text = true] in type {itemType}}}), "Variable X"),
    #"Split Column by Delimiter1" = Table.ExpandListColumn(Table.TransformColumns(#"Split Column by Delimiter", {{"Variable Y", Splitter.SplitTextByDelimiter(",", QuoteStyle.Csv), let itemType = (type nullable text) meta [Serialized.Text = true] in type {itemType}}}), "Variable Y"),
    #"Changed Type" = Table.TransformColumnTypes(#"Split Column by Delimiter1",{{"Index", Int64.Type}, {"Variable X", Int64.Type}, {"Variable Y", Int64.Type}})
in
    #"Changed Type"

Then in DAX you can do a simple Pearson correlation from a quick measure if you like.

, you can modify the auto generated code, or you can roll your own.