## SET UP IF STARTING FROM HERE { setScience(proj="Spotify_Accounting_ETL", subProj="StreamConversion", subl=FALSE) lib(bit64) lib(reshape2) lib(ggplot2) loadIfNotExists("DT.Total_usage") } ## FLAGS spot_using <- "Spotify" spot_using <- "Spotify x-Orch" RatioType_using <- "Ratio_00To30" ### ---------------------------------------------- ### # DT.Total_usage # ------------------------------- # OLD : kCols.DUS <- c("month", "Units", "CutOff") # OLD : kCols.DCgUS <- c("month", "country_code", "Units", "CutOff") # OLD : kCols.DOUS <- c("month", "Orchard_vs_Spotify", "Units", "CutOff") # OLD : kCols.ts <- c("country_code", "CutOff", "Units") # OLD : kCols.US <- c("Units", "CutOff") ## NEW kCols kCols.DS <- c("month", "StreamType") kCols.US <- c( "Units", "StreamType") kCols.DUS <- c("month", "Units", "StreamType") kCols.DO <- c("month", "Orchard_vs_Spotify" ) kCols.DOC <- c("month", "Orchard_vs_Spotify", "country_code", "country_name", "region_group") kCols.DOUS <- c("month", "Orchard_vs_Spotify", "Units", "StreamType") kCols.DCgUS <- c("month", "region_group", "Units", "StreamType") OrchSpotColors <- getDict("Orchard_vs_Spotify") ## MELTED DATA (reshaped in the future) numericCols <- nwhich(canBeNumeric(DT.Total_usage)) DT.Total_usage[, (numericCols) := lapply(.SD, as.numeric), .SDcols = numericCols] DT.Total_usage.melt <- melt(DT.Total_usage, id.vars=c("month", "country_code"), value.name="count", measure.vars=numericCols) ## Shorthand sDT <- DT.Total_usage.melt ## Add in region_group addCountryInfo_(sDT) setcolorderpt(sDT, startCols=c("month", "country_code", "country_name", "region_group")) ## Splice out the column types from the original, unmelted column name sDT[, Orchard_vs_Spotify := ifelse(grepl("orchard", variable), "Orchard", "Spotify")] sDT[, Units := ifelse(grepl("minutes", variable), "Minutes", "Streams")] sDT[Units == "Streams", CutOff := ifelse(grepl("15", variable), 15, ifelse(grepl("30", variable), 30, 0))] ## Calculate the ABSOLUTE stream count (whereas, the orignal had cumsum counts) byCols.abs <- c("month", "country_code", "Units", "Orchard_vs_Spotify") sDT[Units != "Minutes", absolute_count := c(-diff(count), count[CutOff == 30]), by=byCols.abs] # sDT[Units != "Minutes", absolute_count := { # c( # t.00_to_15 = count[CutOff == 0] - count[CutOff == 15] # The latter value includes 30_plus # , t.15_to_30 = count[CutOff == 15] - count[CutOff == 30] # , t.30_plus = count[CutOff == 30] - 0 # no change # )} , by=byCols.abs] ## CONFIRM stopifnot(sDT[Units != "Minutes", absolute_count[CutOff == 0] + count[CutOff == 15] == count[CutOff == 0], by=byCols.abs][, V1]) stopifnot(sDT[Units != "Minutes", absolute_count[CutOff == 15] + absolute_count[CutOff == 30] == count[CutOff == 15], by=byCols.abs][, V1]) stopifnot(sDT[Units != "Minutes", sum(absolute_count) == count[CutOff==0], by=byCols.abs][, V1]) sDT[CutOff == 00, StreamType := "Streams 00-15 Secs"] sDT[CutOff == 15, StreamType := "Streams 15-30 Secs"] sDT[CutOff == 30, StreamType := "Streams 30+ Secs"] sDT[Units == "Minutes", absolute_count := count] sDT[Units == "Minutes", StreamType := "Total Mins Streamed"] sDT[, month := as.Date(month)] ## Remove shorthand { if (identical(DT.Total_usage.melt, sDT)) rm (sDT) else stop ("DT.Total_usage.melt is NOT identical to sDT") } ## Add in "Total Streams" if ("TOTAL STREAMS 0+ Secs" %ni% DT.Total_usage.melt$StreamType) { DT.Total_usage.melt <- rbind( DT.Total_usage.melt , DT.Total_usage.melt[Units != "Minutes", list(StreamType = "TOTAL STREAMS 0+ Secs", CutOff = (-1), absolute_count = count[CutOff==0]), keyby=c(setdiff(names(DT.Total_usage.melt), c("StreamType", "CutOff", "count", "variable", "absolute_count")))] , fill = TRUE) } ## 0 Seconds vs 15 seconds vs 30 Seconds DT.Total_usage.melt.aggdByC <- sumTheTable(DT.Total_usage.melt, byCols=c(kCols.DOUS, "CutOff"), colsUsing="absolute_count") DT.Total_usage.melt.aggdByC[, What := ifelse(grepl("Total", StreamType, ignore.case=TRUE), "Total", "Absolute Counts")] DT.Total_usage.melt.aggdByC[, Units := factor(Units, levels=c("Streams", "Minutes"))] # DT.Total_usage.melt.aggdByC[, Counting := ifelse(CutOff == 0, "00 ~ 15 Seconds [Unpaid]", ifelse(CutOff == 15, " ~ 15 Seconds [Unpaid]", ) ## ADD in Spotify x-Orchard add_in_xOrchard <- function(DT, orchCol="Orchard_vs_Spotify", measureCols=c("absolute_count", "count", "Value", "value"), exclude=c(), xorch_string="Spotify x-Orch", makefactor=TRUE) { calculate_xOrchard <- function(x, OScol) { if (length(x) != 2) warning (sprintf("%s than 2 elements provided to calculate_xOrchard()", ifelse(length(x) < 2, "Less", "More"))) x[OScol == "Spotify"] - x[OScol == "Orchard"] } # browser() tmp_measureCols <- c(orchCol, measureCols) %>% intersect(names(DT)) tmp_dimCols <- setdiff(names(DT), c(tmp_measureCols, exclude)) ## sanity check if (length(tmp_measureCols) == 1 && tmp_measureCols == orchCol) stop ("none of the measureCols where in the DT") tmp_DT.xOrch <- DT[, {.SD; lapply(.SD[, !orchCol, with=FALSE], calculate_xOrchard, .SD[[orchCol]])}, by=tmp_dimCols, .SDcols=tmp_measureCols] %>% cbind(xorch_string) %>% {setnames(., length(.), orchCol)} %>% {.} ## rbind it all together ret <- rbind(DT, tmp_DT.xOrch, use.names=TRUE, fill=TRUE) if (makefactor) ret[, (orchCol) := factor(get(orchCol), levels=c("Orchard", xorch_string, "Spotify"))] return(rbind(DT, tmp_DT.xOrch, use.names=TRUE, fill=TRUE)) } DT.Total_usage.melt %<>% add_in_xOrchard(exclude="variable") ### Question 1: Is the ratio (00-15) / (30+) Constant? DT.ratio_counts <- { DT.Total_usage.melt[!is.na(CutOff) & Orchard_vs_Spotify %in% c("Orchard", spot_using), { ret <- c( Ratio_00To30 = count[CutOff==30] / count[CutOff==00] , Ratio_15To30 = count[CutOff==30] / count[CutOff==15] , Ratio_00To15 = count[CutOff==15] / count[CutOff==00]) list(RatioType=names(ret), Value=ret) } , keyby=kCols.DOC ][RatioType == RatioType_using] } ylims <- c(0.55, 0.85, DT.plot$Value, DT.plot_text_words$y) %>% range # levs.months <- sapply() fmt_month <- "%Y-%b" levs <- DT.ratio_counts[, range(month)] %>% as.list %>% setNames(c("from", "to"))%>% c(by="month") %>% do.call(what=seq.Date, .) %>% format(fmt_month) DT.ratio_counts[, mm := toFactorWithExpectedLevels(format(month, format=fmt_month), levels=levs)] DT.ratio_counts.medians <- DT.ratio_counts[, list(median_value=median(Value)), keyby=list(month, mm, Orchard_vs_Spotify)] DT.ratio_counts.medians_for_segment <- DT.ratio_counts.medians %>% {cbind(.[, head(.SD, -1), keyby=Orchard_vs_Spotify], setnames(copy(.), paste0(names(.), "_end"))[, tail(.SD, -1), keyby=Orchard_vs_Spotify_end])} DT.ratio_counts.means <- DT.ratio_counts[, list(mean_value=mean(Value)), keyby=list(month, mm, Orchard_vs_Spotify)] DT.ratio_counts.means_for_segment <- DT.ratio_counts.means %>% {cbind(.[, head(.SD, -1), keyby=Orchard_vs_Spotify], setnames(copy(.), paste0(names(.), "_end"))[, tail(.SD, -1), keyby=Orchard_vs_Spotify_end])} ## Set the key to OvsS and Ratio Type, for easier filtering later setkeyIfNot(DT.ratio_counts, "Orchard_vs_Spotify", "RatioType", verbose=FALSE) ## TODO: Fix this DT.ratio_counts[, mgroup := ifelse(month <= "2014-05-31", "2014-Feb to 2014-May", sprintf("2014-Jun to %s", mm[month==max(month)][[1]]))] ## Margins for plot margins <- c(top=1, right=2, bottom=1, left=1) * 7 rgrps <- DT.ratio_counts$region_group %>% {if (is.factor(.)) levels(.) else unique(.)} %>% c("WW", .) P.stream_conversions_by_region <- emptylist(rgrps) for (rgrp in rgrps) { ## verbose for user catn("~~~~~~~~~ PLOTTING: ", rgrp, " ~~~~~~~~~~") ## Crop the data to the region group being used DT.ratio_count_for_rgrp <- if (rgrp == "WW") DT.ratio_counts else DT.ratio_counts[(rgrp == region_group)] if (!nrow(DT.ratio_count_for_rgrp)) { message("No rows for '", rgrp, "' -- skipping") next; } ## Words for the graph TITLE <- sprintf("STREAM CONVERSIONS [%s]", ifelse(rgrp=="WW", "World Wide", ifelse(rgrp=="ROW", "Rest of World", rgrp))) %>% subtext("from unmonetized-stream to monetized-stream (ie, longer than 30 seconds)") # ylab <- "Ratio Per Country of [# of streams lasting longer than 30 Seconds] /\n [# of streamss initiated]\n" ylab <- "Ratio Per Country of over(# of streams lasting > 30 Seconds\", \"Total # of streamss initiated)\n" if (!identical(DT.ratio_count_for_rgrp[, unique(RatioType)], "Ratio_00To30")) stop ("This code not implemented for RatioType_using != 'Ratio_00To30'") ## In an older iteration of the code, we were not cropping by RatioType until this point in the code. ## In the current iteration of the code, DT.plot is essentially identical to DT.ratio_count_for_rgrp DT.plot <- copy(DT.ratio_count_for_rgrp[RatioType == RatioType_using]) DT.plot_text_numbs <- DT.plot[(month(month)==5), list(median_avg=fwp(median(Value), 0)), by=list(mm, Orchard_vs_Spotify) ][, y := 0.86 + 0.015 * (-1)^(Orchard_vs_Spotify != "Orchard")] DT.plot_text_words <- DT.plot[(month(month)==3), list(median_avg=sprintf("Median\nMay '%s: ", year(month[[1]]) %% 100)), by=list(mm, Orchard_vs_Spotify) ][, y := 0.86 + 0.015 * (-1)^(Orchard_vs_Spotify != "Orchard")] notch_on <- nrow(DT.plot[(month == max(month)) & Orchard_vs_Spotify == "Orchard"]) > 30 ## Calculate the Averages, used for plotting the horizontal bar (geom_segment) tmp.avgs <- DT.ratio_count_for_rgrp[, median(Value), by=list(Orchard_vs_Spotify, mgroup)] tmp.avgs[, c("xmin", "xmax") := as.data.table(do.call(rbind, strsplit(mgroup, " to ")))] P.stream_conversion <- { ## Both Increased, but Spotify Increased more ggplot() + geom_boxplot(data=DT.plot[.("Orchard")], aes(x=mm, y=Value, fill=Orchard_vs_Spotify, color=Orchard_vs_Spotify), alpha=.40, notch=notch_on) + geom_boxplot(data=DT.plot[.(spot_using)], aes(x=mm, y=Value, fill=Orchard_vs_Spotify, color=Orchard_vs_Spotify), alpha=.40, notch=notch_on) + labs(title=TITLE, y=subtext(main=ylab, subtext=" ")) + scale_x_discrete(breaks=levs[seq(2, length(levs), by=2)]) + geom_point(data=DT.plot[, list(Value, quantile(Value, probs=.75) + 1.5 * IQR(Value)), keyby=c("mm", "Orchard_vs_Spotify")][Value > V2], aes(x=mm, y=Value, color=Orchard_vs_Spotify)) + geom_point(data=DT.plot[, list(Value, quantile(Value, probs=.25) - 1.5 * IQR(Value)), keyby=c("mm", "Orchard_vs_Spotify")][Value < V2], aes(x=mm, y=Value, color=Orchard_vs_Spotify)) + geom_line(data=DT.plot, aes(x=mm, y=median(Value), fill=Orchard_vs_Spotify, color=Orchard_vs_Spotify)) + angledtext(x=20) + percent.y(limits=ylims) + relativetext(x=0.75, y=1.10, what="text") + relativetext(x=1.10, y=0.85, what="title") + legendTitleOff() + legendtop() + theme(plot.margin = unit(margins, "mm")) + ## COLORS color_by_dict(OrchSpotColors) + fill_by_dict(OrchSpotColors) } ## I dont like the way this looks; so removing it for now if (FALSE) { ## Add text for median value this month, and same time last year. geom_text(data=DT.plot_text_numbs, aes(x=mm, y=y, label=median_avg, color=Orchard_vs_Spotify), size=4, fontface="bold") + geom_text(data=DT.plot_text_words, aes(x=mm, y=y, label=median_avg), size=3, color="Black", fontface="bold") } if (rgrp == "WW") { P.stream_conversion <- P.stream_conversion + geom_segment(data=tmp.avgs, aes(x = xmin, xend=xmax, y = V1, yend=V1), color="black" , size=1.5, alpha=0.5) + geom_segment(data=tmp.avgs, aes(x = xmin, xend=xmax, y = V1, yend=V1, color=Orchard_vs_Spotify) , size=1, alpha=1.0) + xlab(subtext("Month", "horizontal line represents the median average for the time frame")) } else { P.stream_conversion <- P.stream_conversion + geom_segment(data = DT.ratio_counts.medians_for_segment, aes(x=mm, y=median_value, xend=mm_end, yend=median_value_end, color=Orchard_vs_Spotify)) + xlab(subtext("", "horizontal line represents the *worldwide* median average")) } P.stream_conversions_by_region[[rgrp]] <- P.stream_conversion } ## SAVE ## For a single plot: # ggsave.out(P.stream_conversion, open=TRUE, width=8, height=8) ## For multiple plots: P.stream_conversions_by_region %<>% removeNullsFromList() printToPDF(P.stream_conversions_by_region, height.per.plot=9, width.per.plot=8, paginate=TRUE, open=TRUE) print(ggsave.out(plot=P.stream_conversions_by_region[["WW"]], nm="Stream Conversions Worldwide", ext="png")) "&&& LEFT OFF HERE .... " "TODO :: marrange to group of 3" # | # |## Different Plot type # |# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ # # | DT.Total_usage.melt.aggdByC %<>% add_in_xOrchard() # | P <- # | { # | ggplot(DT.Total_usage.melt.aggdByC, aes(x=month, y=absolute_count, color=StreamType)) + # | geom_line(aes(linetype=Units), size=1.4) + # | facet_grid(What + Orchard_vs_Spotify ~ ., scale="free_y") + # | millions.y() + ylab("Count in Millions") + # | ggtitle("Usage by Month (Minutes or Streams) -- Orchard vs Spotify") # | } # | # | P2 <- colorByGroup(P, OrchSpotColors, showWarnings=FALSE) # | plot(P2) # |# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ # # | # | # | # | # | # |AFTER THIS IS SCRATCH WORK FROM A YEAR AGO... # |CHECK IF IT HAS ANYTHING MEANINGUFL, THEN KILL IT # | # |# This whole file used to be called # |# ~/git/orch/src/Spotify_Accounting_ETL/02b DT.data.Total_usage - absolute.r # |# And relates to # |# ~/git/orch/src/Spotify_Accounting_ETL/02 Ingest Spotify Files.r # |# etc. # | # | # | # | # | # | # | # | # |------------------------------- # |## update 2015-08-27 # |## I am not sure what the purpose to DT.ratio_counts2 was? # |## I was comparing it to DT.ratio_counts for some reason # |if (FALSE) # |{ # | DT.ratio_counts2 <-DT.Total_usage[, list(Orchard_Ratio_00To30 = orchard_streams_above_30_sec / orchard_streams # | , Spotify_Ratio_00To30 = all_streams_above_30_sec / all_streams) # | , keyby=list(month, country_code)] # | DT.ratio_counts2 <- melt(DT.ratio_counts2, id.vars=key(DT.ratio_counts2)) # | DT.ratio_counts2[, Orchard_vs_Spotify := ifelse(grepl("Orchard", variable), "Orchard", "Spotify")] # | DT.ratio_counts2[, variable := gsub(".*\\.", "", variable)] # | setnames(DT.ratio_counts2, "variable", "RatioType") # | setnames(DT.ratio_counts2, "value", "Value") # | matchKey(DT.ratio_counts, DT.ratio_counts2, kCols.DOC, organize=TRUE) # | identical(DT.ratio_counts[RatioType == "Ratio_00To30"]$Value, DT.ratio_counts2$Value) # |} # | # |------------------------------- # |### -------- PAID VS UNPAID -------------------- ### # |## START BY dcasting # |DT.Total_usage.dcast <- dcast.data.table(DT.Total_usage.melt[TRUE][,StreamType := ifelse(grepl("15", StreamType), "Streams 00-30 Secs", StreamType)], month + Units + StreamType ~ Orchard_vs_Spotify, value.var="absolute_count", fun.aggregate=sum) # |DT.Total_usage.dcast <- DT.Total_usage.dcast[!grepl("Total Mins", StreamType)] # |DT.Total_usage.dcast[, StreamType := ifelse(grepl("00", StreamType), "Unpaid Streams", ifelse(grepl("TOTAL", StreamType), "TOTAL Streams", "Paid Streams"))] # |DT.Total_usage.dcast[, StreamType := factor(StreamType, levels=c("Unpaid Streams", "Paid Streams", "TOTAL Streams"))] # |setkeyIfNot(DT.Total_usage.dcast, kCols.DUS, verbose=FALSE, organize=TRUE) # | # |## Add ratio column # |DT.Total_usage.dcast[, orch_to_spot_ratio := Orchard / Spotify] # | # |### ------------------- # |TITLE <- "Number of Stream (or Total Minutes)\nComparing Orchard to Spotify, by Month and Stream cut-off" # |P.orchard_to_spotify_ratio_paid_unpaid <- # |ggplot(DT.Total_usage.dcast, aes(x=month, y=orch_to_spot_ratio, color=StreamType)) + geom_line(size=1, aes(linetype=StreamType)) + geom_point(alpha=.3, size=2.5) + facet_grid(Units ~ ., scale="free_y") + percent.y() + labs(y="Ratio of Orchard to Spotify, by group\n", title=subtext("Orchard Streams as a Percentage of Spotify Streams", "NOTE: Orchard includes RED")) + scale_linetype_manual(values=c(1, 1, 2)) # |ggsave.out(P.orchard_to_spotify_ratio_paid_unpaid, open=TRUE) # |### ----------------------------------------------------------- # |------------------------------- # | # | # |## START BY dcasting # |DT.Total_usage.dcast <- dcast.data.table(DT.Total_usage.melt, month + Units + StreamType ~ Orchard_vs_Spotify, value.var="absolute_count", fun.aggregate=sum) # | # |setkeyIfNot(DT.Total_usage.dcast, kCols.DUS, verbose=FALSE, organize=TRUE) # | # |## Add ratio column # |DT.Total_usage.dcast[, orch_to_spot_ratio := Orchard / Spotify] # |DT.Total_usage.dcast[, DELTA_orch_to_spot_ratio := diffNA(orch_to_spot_ratio), by=kCols.US] # | # |## Percent of increase # |DT.Total_usage.dcast[, Orch.perc_change := percentIncrease(Orchard), by=kCols.US] # |DT.Total_usage.dcast[, Spot.perc_change := percentIncrease(Spotify), by=kCols.US] # |DT.Total_usage.dcast[, DELTA_Orch.perc_change := diffNA(Orch.perc_change), by=kCols.US] # |DT.Total_usage.dcast[, DELTA_Spot.perc_change := diffNA(Spot.perc_change), by=kCols.US] # | # |## SEE # |DT.Total_usage.dcast[.(as.Date(c("2014-04-01", "2014-05-01")))][order(StreamType)] # |DT.Total_usage.dcast[.(as.Date(c("2014-04-01", "2014-05-01")))][order(StreamType)] # | # | # |TITLE <- "Number of Stream (or Total Minutes)\nComparing Orchard to Spotify, by Month and Stream cut-off" # |## OLD # |ggplot(DT.Total_usage.dcast, aes(x=month, y=orch_to_spot_ratio, color=StreamType)) + geom_line(size=1) + geom_point(alpha=.3, size=2.5) + facet_grid(Units ~ ., scale="free_y") + percent.y() # | # |## same, no minutes # |### ------------------- # |P.orchard_to_spotify_ratio <- ggplot(DT.Total_usage.dcast[Units != "Minutes"], aes(x=month, y=orch_to_spot_ratio, color=StreamType)) + geom_line(size=1) + geom_point(alpha=.3, size=2.5) + facet_grid(Units ~ ., scale="free_y") + percent.y() + labs(y="Ratio of Orchard to Spotify, by group\n", title=subtext("Orchard Streams as a Percentage of Spotify Streams", "NOTE: Orchard includes RED")) # |ggsave.out(P.orchard_to_spotify_ratio) # |### ------------------- # | # | # |ggplot(DT.Total_usage.dcast[Units != "Minutes"], aes(x=month, y=DELTA_orch_to_spot_ratio, color=StreamType)) + geom_line(size=1) + geom_point(alpha=.3, size=2.5) + percent.y() + geom_hline(y=0, color="white", linetype=1, size=1.2, alpha=.8) + geom_hline(y=0, color="black", size=.75, linetype=4, alpha=.6) # | # |## NEW 20141020 # |ggplot(DT.Total_usage.dcast, aes(x=month, y=orch_to_spot_ratio, color=StreamType)) + geom_line(size=1) + geom_point(alpha=.3, size=2.5) + percent.y() + labs(title=TITLE, y="Orchard as a percent of Spotify") # | # |ggplot(DT.Total_usage.dcast, aes(x=month, y=DELTA_orch_to_spot_ratio, color=StreamType)) + geom_line(size=1) + geom_point(alpha=.3, size=2.5) + percent.y() + labs(title=paste0("CHANGE IN ", TITLE), y="CHANGE IN Orchard as a percent of Spotify") # | # |ggplot(DT.Total_usage.dcast, aes(x=month, y=DELTA_Orch.perc_change, color=StreamType)) + geom_line(size=1) + geom_point(alpha=.3, size=2.5) + percent.y() + labs(title=paste0("CHANGE IN ", TITLE), y="CHANGE IN Orchard as a percent of Spotify") # |dev.new() # | # |--------------------------------------- # |## Same as above but with country_code added # |## START BY dcasting # |DT.Total_usage.dcast <- dcast.data.table(DT.Total_usage.melt, month + country_code + Units + CutOff ~ Orchard_vs_Spotify, value.var="count", fun.aggregate=sum) # | # |setkeyIfNot(DT.Total_usage.dcast, kCols.DCgUS, verbose=FALSE, organize=TRUE) # | # |## Add ratio column # |DT.Total_usage.dcast[, orch_to_spot_ratio := Orchard / Spotify] # | # |## SEE # |DT.Total_usage.dcast[.(as.Date("2014-07-01"), "US")] # | # |## Scale Uniform to [0, 1] # |# DT.Total_usage.dcast[, Orch.scaled_01 := scaleunif(Orchard), by=kCols.ts] # |# DT.Total_usage.dcast[, Spot.scaled_01 := scaleunif(Spotify), by=kCols.ts] # | # |## Scale to SD # |DT.Total_usage.dcast[, Orch.scaled := scale(Orchard), by=kCols.ts] # |DT.Total_usage.dcast[, Spot.scaled := scale(Spotify), by=kCols.ts] # | # |## Percent of increase # |DT.Total_usage.dcast[, Orch.perc_change := percentIncrease(Orchard), by=kCols.ts] # |DT.Total_usage.dcast[, Spot.perc_change := percentIncrease(Spotify), by=kCols.ts] # | # |## SEE # |DT.Total_usage.dcast[.(as.Date(c("2014-04-01", "2014-05-01")), "US")][order(CutOff)] # |DT.Total_usage.dcast[.(as.Date(c("2014-04-01", "2014-05-01")), "UY")][order(CutOff)] # | # | # |ggplot(DT.Total_usage.dcast, aes(x=month, y=orch_to_spot_ratio, color=factor(CutOff))) + geom_line() + facet_grid(Units ~ ., scale="free_y") # | # |--------------------------------------- # |kCols.region_group <- gsub("^country_code$", "region_group", kCols.DCgUS) # |DT.plot.agg <- melt(DT.Total_usage.dcast[!is.na(CutOff) & CutOff == 30], id.vars=kCols.DCgUS)[, variable := as.character(variable)] # |DT.plot.agg[, Orchard_vs_Spotify := ifelse(grepl("orch", as.character(variable), ignore.case=TRUE), "Orchard", "Spotify")] # |addColsFrom_(DT.plot.agg, DT.country, joinCols="country_code") # |DT.plot.agg <- DT.plot.agg[region_group != "Canada"] # |DT.plot.agg <- DT.plot.agg[grepl("perc_change$", variable), lapply(.SD, sumn), keyby=c(kCols.region_group, "Orchard_vs_Spotify"), .SDcols="value"] # | # |DT.plot.agg[.(as.Date("2014-04-01"))] # | # |y.var <- "value" # |ggplot(data = DT.plot.agg[month > min(month)], aes_string(x="month", y=y.var, color="region_group", linetype="Orchard_vs_Spotify")) + geom_line() + geom_point(alpha=.5) + ylab(paste0(gsub("_", " ", y.var), " (percent)")) # | # |+ facet_grid(type_of_trial ~ ., scales="free_y") # |--------------------------------------- # | # |DT.plot.agg <- melt(DT.Total_usage.dcast[!is.na(CutOff) & CutOff == 30], id.vars=kCols.DCgUS)[, variable := as.character(variable)] # |DT.plot.agg[, Orchard_vs_Spotify := ifelse(grepl("orch", as.character(variable), ignore.case=TRUE), "Orchard", "Spotify")] # |addColsFrom_(DT.plot.agg, DT.country, joinCols="country_code") # | # |y.var <- "value" # |ggplot(data = DT.plot.agg[grepl("scaled_01", variable)], aes_string(x="month", y=y.var, color="region_group", linetype="Orchard_vs_Spotify")) + geom_line() + geom_point(alpha=.5) + ylab(paste0(gsub("_", " ", y.var), " (percent)")) # | # |+ facet_grid(type_of_trial ~ ., scales="free_y") # |--------------------------------------- # | # | # |whose <- c(orchard="orchard_streams", all="all_streams") # |length <- c(touched="", short="_above_15_sec", full="_above_30_sec") # | # |## ------------------------------------ ## # |## STILL TODO ## # |## ------------------------------------ ## # | for (l in length) # | DT.Total_usage[, paste0("orch_to_all_ratio", l) := get(paste0("orchard_streams", l)) / get(paste0("all_streams", l))] # | DT.Total_usage[, orch_to_all_ratio_minutes := orchard_minutes_streamed / minutes_streamed] # | ## # | DT.Total_usage[, perc.songs_skipped := 1 - all_streams_above_30_sec/all_streams] # | DT.Total_usage[, perc.orchard_songs_skipped := 1 - orchard_streams_above_30_sec/orchard_streams] # |## ------------------------------------ # | # |addColsFrom_(DT.Total_usage, DT.country, joinCols="country_code") # | # |------------------------------------------------ # | # | # |plotSpotifyData <- function(DT, y.vars=c("perc.songs_skipped")) # |## Look at USA vs ROW # |DT.plot <- copy(DT.Trial_uptake) # |y.var = "No._of_trials_starting_during_the_month" # | # |addColsFrom_(DT.plot, DT.country, joinCols="country_code") # | # |byCols <- c(setdiff(key(DT.plot), "country_code"), "region_group") # |DT.plot[, country_code := NULL] # | # | # |## Aggregate # |DT.plot.agg <- DT.plot[, lapply(.SD, sumn), keyby=byCols] # |## Drop any type that only has one or two data points # |DT.plot.agg <- DT.plot.agg[type_of_trial %in% DT.plot.agg[, .N, by=type_of_trial][N > 2, type_of_trial]] # | # |lib(ggplot2) # |ggplot(data = DT.plot.agg, aes_string(x="month", y=y.var, color="region_group")) + geom_line() + facet_grid(type_of_trial ~ ., scales="free_y") + millions.y() + ylab(paste0(gsub("_", " ", y.var), " (in millions)")) # |----------------------------