tslumley

library(hextri)
library(survey)
data(api)

## all hexes same orientation: looks as if middle schools went up more
with(apipop,hextri(api99,api00,stype,c("orange","forestgreen","purple"),nbins=20,diffuse=FALSE,
   xlab="1999 Academic Performance Index",ylab="2000 Academic Performance Index",style="size"))
legend("topleft",fill=c("orange","purple","forestgreen"),legend=c("Elementary","Middle","High"),bty="n")
 
## random orientation: less pretty, but you can see middle schools went up less

tslumley

rock<-read.csv("~/Downloads/output.csv", as.is=TRUE, header=FALSE, col.names=c("datetime","song","group"))

rock$stime<-substr(rock$datetime,5,9)

rock$timeangle<-with(rock, as.numeric(substr(stime,1,2))*2*pi/24+as.numeric(substr(stime,4,5))*2*pi/24/60)

rock$ty<-cos(rock$timeangle)
rock$tx<-sin(rock$timeangle)
rock$txmean<-mean(rock$tx)
rock$tymean<-mean(rock$ty)

tslumley

hexmap<-function(x,y,id,...){
	xcent<-x*sqrt(3) - (y%%2) *sqrt(3)/2
	ycent<-y*1.5
	
	plot(x,y,type="n",ylim=range(ycent)+c(-2,2),xlim=range(xcent)+c(-2,2))
	for(i in 1:length(x)){
		polygon(hex_x+xcent[i],hex_y+ycent[i])
		text(xcent[i],ycent[i],id[i],cex=0.4)
	}

tslumley

Who gets interrupted on the Supreme Court

Data from Adam Feldman's Empirical SCOTUS blog: https://empiricalscotus.com/2016/05/26/interruptions/ https://empiricalscotus.com/2016/05/01/oral-arguments-2015/

as analysed at StatsChat: http://www.statschat.org.nz/2016/05/29/ima-let-you-finish/

tslumley

Data from the New Zealand Department of Corrections, provided to Mark Hanna under the Official Information Act

Original documents available at https://fyi.org.nz/request/3926-strip-searches-in-new-zealand-prisons#incoming-13386

tslumley

gravity<-mks(c(3.7,8.9,9.8,3.7,23.1,9.0,8.7,11.0,0.6), m=1, kg=0, s=-2)
density<-mks(c(5427,5243,5515,3933,1326,687,1270,1638,2390), m=-3,kg=1,s=0)

speed.of.light<-fff(1.8026175e12,fur=1,fir=0,ftn=-1)

setMethod("mean","mks",function(x,...){
	u<-getUnits(x)
	mks(mean(getValues(x)),u[1],u[2],u[3])
	})

tslumley

"+.bentime"<-function(e1,e2){
	   e<-list(hour=e1$hour+e2$hour,min=e1$min+e2$min,sec=e1$sec+e2$sec)
	   sec_overflow<-e$sec>60L
	   e$sec<-e$sec %% 60L
	   e$min<-e$min+sec_overflow
	   min_overflow<-e$min>60L
	   e$hour<-e$hour+min_overflow
	   e$min<-e$min %% 60L
	   class(e)<-"bentime"

tslumley

pal<-function(mat) rgb(mat[,1],mat[,2],mat[,3],maxColorValue=255)


showpal<-function(pal) {
	n<-length(pal)
	image(1:n, 1, as.matrix(1:n), col = pal, 
        xlab = deparse(substitute(pal)), ylab = "", xaxt = "n", yaxt = "n", 
        bty = "n")
}

tslumley

This is a model for NZ road deaths per billion km travelled, using data complied by Sam Warburton (@economissive on Twitter)

The model has an underlying smoothed trend which is a random walk with Gaussian increments, and a year-specific deviation that is also Gaussian. That is, it's a state space model with a random walk in the latent variable and a measurement model that's a Poisson-logNormal mixture to get overdispersed counts.

The variances of the random walk steps and the year-specific deviations have diffuse hyperpriors and the means are set up so that the mean of the exponential of the increment is 1 -- ie, so mu is genuinely the trend in means.

tslumley

An example of the non-transitivity of the Wilcoxon test, using Efron's non-transitive dice.

refron() generates numbers from a set of four non-transitive dice, so that a beats b beats c beats d beats a two-thirds of the time, with optional Gaussian smoothing.

example.R does Wilcoxon rank-sum tests to show that a > b > c > d > a