ordBTL (version 0.8)

design: Design Matrix

Description

This function returns the design matrix for an ordinal Bradley-Terry-Luce model.

Usage

design(X, var1, var2, use.vars=NULL, reference=NULL, prefix="GAMMA", prefix.home="ALPHA", home.advantage=c("no","specific","yes"))

Arguments

X

a data frame in long format (see wide2long).

var1

a character of the column name from X specifying the first object to be compared (or, in sport context, the home team).

var2

a character of the column name from X specifying the second object to be compared (or, in sport context, the away team).

use.vars

a character vector with the names of additional covariates of X that will additionally be included into the design matrix. (example: use.vars = c("ENG", "SEX")) if the covariates ENG and SEX should be included. If all covariates of X should be included, you can use use.vars = "ALL". The default is use.vars = NULL for no additional covariates.

reference

a character specifying the reference object.

prefix

(optional) a character added in the names of the estimated object parameters

prefix.home

(optional) a character added in the names of the estimated home advantage parameters

home.advantage

Note that the home advantage is equivalent to an order effect home.advantage="no" uses no home advantage (order effect), home.advantage="specific" uses one home advantage (order effect) for each object and home.advantage="yes" uses one home advantage (order effect) for any object.

Value

A data frame where each row refers to a pair comparison and each column corresponds to an object.

Examples

# load german football-league (Bundesliga) data
library("wikibooks")
data(Bundesliga)

# add new variable Y3 reflecting the response which is coded as 
# 1 if the home team wins
# 2 if the game ends up with a tie
# 3 if the home team loses
diff <- Bundesliga$Tore.Heim - Bundesliga$Tore.Gast
Bundesliga$Y3 <- as.ordered(ifelse(diff >= 1, 1, 
                                   ifelse(diff <= -1, 3, 2)))
buli0506 <- subset(Bundesliga, Saison=="2005/2006")
str(buli0506)

# Design matrix without home advantage
des.nohome <- design(buli0506, var1="Heim", var2="Gast", 
                     home.advantage="no")
str(des.nohome)

# Design matrix with one home advantage parameter for all objects
des.onehome <- design(buli0506, var1="Heim", var2="Gast", 
                      home.advantage="yes")
str(des.onehome)

# Design matrix with home advantage parameters for each object
des.teamhome <- design(buli0506, var1="Heim", var2="Gast",
                       home.advantage="specific")
str(des.teamhome)

# Design matrix with additional covariable "Spieltag"
des.covs <- design(buli0506, var1="Heim", var2="Gast", 
                   use.vars=c("Spieltag"), home.advantage="no")
str(des.covs)