udd-smooth-ai.R

#
# udd-smooth-ai.R, 22 Apr 20
#
# Data from:
# Impact of Installation Counts on Perceived Quality: A Case Study of Debian
# Israel Herraiz and Emad Shihab and Thanh H. D. Nguyen and Ahmed E. Hassan
#
# Evidence-based Software Engineering: based on the publicly available data
# Derek M. Jones
#
# Example from:
# Evidence-based Software Engineering: based on the publicly available data
# Derek M. Jones
#
# TAG Debian_package package_install package_age


source("ESEUR_config.r")


library("MASS")


plot_layout(3, 1, max_height=ESEUR_max_height-2)
par(mar=MAR_default-c(0.6, 0.0, 1.0, 0.0))


q1=read.csv(paste0(ESEUR_dir, "regression/Q1_udd.csv.xz"), as.is=TRUE)
q10=read.csv(paste0(ESEUR_dir, "regression/Q10_udd.csv.xz"), as.is=TRUE)

udd=merge(q1, q10)

plot(udd$age, udd$insts, log="y", col=point_col,
	xaxs="i",
	cex.axis=1.5, cex.lab=1.5,
	xlab="Age (days)", ylab="Installations\n")
smoothScatter(udd$age, log(udd$insts),
	xaxs="i",
	cex.axis=1.5, cex.lab=1.5,
	xlab="Age (days)", ylab="log(Installations)\n")

plot(udd$age, udd$insts, log="y", col=point_col,
	xaxs="i",
	cex.axis=1.5, cex.lab=1.5,
	xlab="Age (days)", ylab="Installations\n")

# There is no log option, so we have to compress/expand ourselves
d2_den=kde2d(udd$age, log(udd$insts+1e-5), n=50)
contour(d2_den$x, exp(d2_den$y), d2_den$z, nlevels=5, add=TRUE)