Differential analytics download

asciimath.txt

Abbreviations
----
v value
exp experiment
c contrast 
id identifier
as analytics search, which the query either matches or it doesn't
cs conditions search, which the query either matches or it doesn't

Sets considered
----
D= {(v,exp,c,id)} all data we have
A= {(as,v, exp,c,id): (v,exp,c,id) is in D, v>threshold, as=annotateAll(v,exp,c,id)}.
C= {(cs, exp,c): (v,exp,c) is in D for some v, cs= annotateConditions(exp,c)}
So as and cs are calculated fields

Current search
----
R = { id: there is (as,v, exp, c, id) in A such that as matches query} 
    x {(exp, c): there is (cs,exp, c) in C such that cs matches query } intersect D

Search in analytics index only
----
S = { (v,exp,c,id) : there is (as, v, exp,c,id) in A such that as matches query}


What data will we miss?
----
Consider (exp', c', id') in R\S.

If there are (as,v) such that (as, v, exp', c', id') is in A, then as doesn't match query. 
Hence for all (as, exp'', c'', id') in A such that as matches query, (exp'',c'') != (exp', c')

Hence there are some (as*,cs*): 
(as*, v, exp'', c'', id') is in A
(cs*,exp', c') is in C
as* matches query and cs* matches query
exp'' is not exp' or exp'' is exp' but c'' is not c'
I claim we don't want these results.