DT.EU_aggd_reshaped
"I want to compare the delta from Wk1-Pre  relative to the delta Pre-PrePre"



DT.EU_aggd.delta <- {
  do.call(data.table:::merge.data.table,
    c(lapply(sprintf("delta%s", countCols), function(what) {
      setkeyIfNot(
        data.table:::melt.data.table(DT.EU_aggd_reshaped, measure.vars=extract(what, DT.EU_aggd_reshaped), value.name=what, id.vars=key(DT.EU_aggd_reshaped), variable.name="TimePeriod")[, TimePeriod := gsub(paste0(what, "\\."), "", TimePeriod)] 
        , c(key(DT.EU_aggd_reshaped), "TimePeriod"), verbose=FALSE)
    }), all=TRUE, allow=TRUE)
  )
}

key.bak <- key(DT.EU_aggd.delta)
DT.EU_aggd.delta[, median_deltaunits.byGrp := median(deltaunits), by=setdiff(key(DT.EU_aggd.delta), upcCols)]
DT.EU_aggd.delta[, median_deltarevenue.byGrp := median(deltarevenue), by=setdiff(key(DT.EU_aggd.delta), upcCols)]
DT.EU_aggd.delta[, aboveMedianUnits := deltaunits > median_deltaunits.byGrp]


## Clean up output text
DT.EU_aggd.delta[, TreatmentWk1 := {levels(TreatmentWk1) <- paste(sprintf("%s (%s)", c("No Change", "Increase +1", "Increase +2"), levels(TreatmentWk1))); TreatmentWk1} ]
DT.EU_aggd.delta[TimePeriod == "pre_minus_prePre", TimePeriod := "Before_Experiment"]
DT.EU_aggd.delta[TimePeriod == "wk1_minus_pre",    TimePeriod := "During_Experiment"]

addColsFrom_(DT.EU_aggd.delta, DT.upc_has_override, joinCols=intersect(c("upc", "country_code"), names(DT.upc_has_override)), colsToBring=upc_addl_metaCols)

tmp.newKey <- c(setdiff(key.bak, "TimePeriod"), upc_addl_metaCols, "TimePeriod")
setkeyIfNot(DT.EU_aggd.delta, tmp.newKey, organize=TRUE, verbose=FALSE)


# DT.EU_aggd.delta[, lapply(.SD, median, na.rm=TRUE), keyby=list(store_name, TimePeriod, country_code, category, transac_type_abbr, TreatmentWk1), .SDcols=c("deltarevenue", "deltaunits")][country_code==ctry]
# DT.EU_aggd.delta[(upc == upc[[1]])]
# DT.EU_analytics[upc == DT.EU_aggd.delta[1, upc]]

# store <- "iTunes"
# transac <- "DA"
# ctry <- "GB"

# deltarevenue

# P.deltaunits <- lapply(countries_using, function(ctry)
#   ggplot(data=DT.EU_aggd.delta[store_name == store][country_code == ctry], aes(x=TimePeriod, y=deltarevenue, color=daily_avg_album_units_30preTest)) + geom_jitter(width=0.05) + geom_boxplot(alpha=.85) + facet_grid("category ~ TreatmentWk1 ") + ylim(-20, 20) + ggtitle(paste("Change in ")) + legendtop(TRUE)
#   )

#   ggplot(data=DT.EU_aggd.delta[store_name == store][country_code == ctry][transac_type_abbr == transac]) + 
#   aes(x=daily_avg_album_units_30preTest, y=deltaunits, color=TreatmentWk1) +
#    geom_jitter(width=0.15, alpha=0.35) +
#    log.x() + 
#    ggtitle(sprintf("%s - %s - %s", store, transac, DT.country[.(ctry)]))

# CALCULATE REVENUE WOULD HAVE GENERATED

#    facet_grid("category ~ TreatmentWk1 ") +
#    ylim(-20, 20) +
#    ggtitle(paste("Change in ")) +
#    legendtop(TRUE)