In RobinHankin/hyper2: The Hyperdirichlet Distribution, Mark 2

set.seed(0)
knitr::opts_chunk$set(echo = TRUE)
library("hyper2")
library("magrittr")
options("digits" = 5)

knitr::include_graphics(system.file("help/figures/hyper2.png", package = "hyper2"))

To cite the hyper2 package in publications, please use @hankin2017_rmd. Rank tables can be confusing. The package includes a range of functionality for dealing with them, including a bespoke print method. Here, I discuss two cases: firstly, a tiny toy example, and secondly the volvo ocean racing dataset.

Toy dataset

OT <- cbind(1:5,c(1,5,2:4))
rownames(OT) <- letters[seq_len(nrow(OT))]
colnames(OT) <- c("J1","J2")
OT

Object OT is a straightforward order table, just like skating_table but smaller and simpler. We have two columns, one for each judge (named J1 andJ2), and five competitors, a through e. The first row refers to competitor a; both judges put competitor a first. The second row shows that J1 put b second and J2 put competitor b fifth. If instead we ask who was placed third by each judge, this is not directly given; but by looking for 3 in the table we see that J1 put c third and J2 put d third. To answer this and similar questions directly, we can coerce order table OT to a rank table RT:

RT <- ordertable_to_ranktable(OT)
RT

Object RT gives the same information as OT but in a different form. Note that above we see the print method in action, which uses ranktable_to_printable_object(). Here the rows correspond to the two judges. Each row is a rank vector. Column headings refer to placings (c1 is supposed to be read "came first", c2 "came second" etc). Looking at the fifth column (c5) we see that J1 and J2 placed e and b fifth respectively.

Note that it is possible to view object RT directly by removing class ranktable:

class(RT)
class(RT) <- "matrix"  # removes "ranktable" from class list
RT

Note carefully that the second row of RT (which corresponds to the second judge) does not match the second column of OT. The way to read the second row of RT is as follows. We see the entries 1 3 4 5 2 indicating that the first choice was competitor 1 [viz a], the second choice was competitor 3 [viz c], the third choice was competitor 4 [viz d], fourth choice was competitor 5 [e] and fifth choice was competitor 2 [b]. So the rows are ranks (see the discussion in rrank.Rd). But it is very tempting to read the second row of RT as "judge 2 placed competitor b in third place", which is wrong because this is interpreting the row as an order. This is why the print method is there.

Volvo ocean racing dataset

(See volvo.Rd for more details).

volvo_table

Thus we see from the first column that in leg 1, AbuDabhi came first, Brunel third, and so on. From the first row we see that AbuDhabi came first in leg 1, third in leg 2, second in leg 3, and so on. Suppose we wish to know who came first in each leg:

x <- wikitable_to_ranktable(volvo_table)
x

And we see from the first column that AbuDhabi came first in leg 1, Brunel came first in leg 2, Dongfeng came first in leg 3, and so on. The second column shows who came second in each leg. Now suppose we are interested in only the first, second and third-placed teams:

x[,1:3]

The above shows that the extraction method for ranktable objects is not yet written; maybe one day I will get round to implementing [.ranktable() (right now I am not 100% sure that this is a good idea). Note that simply restoring the ranktable class can be misleading:

xx <- x[,1:3]
xx
class(xx) <- "ranktable"
xx

In the above, NAs are introduced because a ranktable object uses the column names to store names of the competitors, and of course extracting the first three columns means keeping only the first three column names. Observe that extraction after passing to ranktable_to_printable_object() works nicely: