(Note- click on graphics for full-size versions)
> topArtistFrame <- subset(factorized, Sampling.Artist %in% SAFreq[1:10,1])
> m <- ggplot(topArtistFrame,aes(y=Sampled.Publishing.Date,x=Sampling.Artist[,drop=TRUE])) + scale_y_continuous(limits=c(1950,1990)) + scale_x_discrete("",breaks=NA)
> m + geom_point(aes(size=Sampled.Song.Frequency,colour=Sampled.Genre),position=position_jitter(height=.2,width=.35),alpha=I(.6))+ stat_boxplot(colour="grey50",alpha=0,outlier.size=1) + geom_text(aes(y=Sampled.Publishing.Date,label=Sampling.Artist[,drop=TRUE]),colour="white",size=3) + labs(y="Sampled Publishing Date",colour="Sampled Genre",size="Sampled Song Frequency") + coord_flip()
+ opts(title="Sampling Distributions For Top 10 Sampling Artists")
> m <- ggplot(factorized,aes(x=Sampling.Artist.Frequency,fill=Sampled.Genre))
> m + stat_density() + scale_y_continuous(name="",breaks=NA) + labs(x="# of Total Samples Per Artist",fill="Sampled Genre")+ opts(title="Density of Samples by Total Samples Per Sampling Artist")
> m <- ggplot(factorized,aes(x=Sampled.Song.Frequency,fill=Sampled.Genre))
> m + stat_density() + scale_y_continuous(name="",breaks=NA) + scale_x_continuous(limits=c(0,15)) + labs(x="# of Total Samples Per Sampled Song",fill="Sampled Genre") + opts(title="Density of Samples by Samples Per Sampled Song")
These series of graphs portray several basic conclusions, extrapolated out from our initial exploratory graphs. The main graph of the top 10 sampling artists shows the level of genre sample diversity they represent, while, as can be seen by the size of each dot, representing sampling frequency, our Funk etc. category is clearly repeated sampled wherever it is used. Other conclusions from the main chart:
- Many of the frequently sampled songs are in evidence across the top artists.
- The top artists who are more popular/mainstream tend to have a larger box boarder than smaller, less popular acts- despite sampling frequency. See Ice Cube vs. DJ Shadow.
- Despite whatever individual choices regarding samples, almost all the artists still cluster at the same time frame.
The other two companion charts back up our major observations. When observing density as a function of sampling artist, as the sampling artist becomes more prolific, the genres begin to even out in distribution. In comparison, as the frequency of sampled songs increases, the Funk Etc. category generally holds steady (note: some Funk category outliers were eliminated which went almost to 100 samples).