genwr: General Win Loss Statistics
In WWR: Weighted Win Loss Statistics and their Variances
Description
Usage
Arguments
Details
Value
Author(s)
References
See Also
Examples
Description
Calculate the general win loss statistics and their corresponding variances under the global NULL hypothesis and under alterantive hypothesis based on Bebu and Lachin (2016) paper, which is a generalization of the win ratio of Pocock et al. (2012) and the win difference of Luo et al. (2015). This calculation needs the users to specify the win loss matrix.
Usage
1
genwr(aindex)
Arguments
aindex
a numeric matrix of win loss indicators. Suppose there are group 1 and group 0 in the study with sample sizes n_1 and n_0 respectively. The matrix aindex is a n_1\times n_0 matrix with elements C_{ij}: i=1,…,n_1, j=1,…,n_0. The element C_{ij} is equal to 1 if subject i in group 1 wins over subject j in group 0 on the most important outcome, C_{ij} is equal to -1 if subject i in group 1 loses against subject j in group 0 on the most important outcome; C_{ij} is equal to 2 if subject i in group 1 wins over subject j in group 0 on the second most important outcome after tie on the most important outcome, C_{ij} is equal to -2 if subject i in group 1 loses against subject j in group 0 on the second most important outcome after tie on the most important outcome;
C_{ij} is equal to 3 if subject i in group 1 wins over subject j in group 0 on the third most important outcome after tie on the first two most important outcomes, C_{ij} is equal to -3 if subject i in group 1 loses against subject j in group 0 on the third most important outcome after tie on the first two most important outcomes; and so forth until all the outcomes have been used for comparison; then C_{ij} is equal to 0 if an ultimate tie is resulted.
Details
General win loss statistics
Value
n1
Number of subjects in group 1
n0
Number of subjects in group 0
n
Total number of subjects in both groups
totalw
Total number of wins in group 1
totall
Total number of losses in group 1
tw
A vector of total numbers of wins in group 1 for each of the outcomes. Note that totalw=sum(tw), the first element is for the most important outcome, the second elemnet is for the second important outcome etc.
tl
A vector of total numbers of losses in group 1 for each of the outcomes. Note that totall=sum(tl), the first element is for the most important outcome, the second elemnet is for the second important outcome etc.
xp
The ratios between tw and tl
cwindex
The win contribution index defined as the ratio between tw and totalw+totall
clindex
The loss contribution index defined as the ratio between tl and totalw+totall
wr
Win ratio defined as totalw/totall
vr
Asymptotic variance of the win ratio under alterantive hypothesis
vr0
Asymptotic variance of the win ratio under global null hypothesis
tr
standardized log(wr) using the variance vr
pr
2-sided p-value of tr
tr0
standardized log(wr) using the variance vr0
pr0
2-sided p-value of tr0
wd
Win difference defined as totalw-totall
vd
Asymptotic variance of the win difference under alterantive hypothesis. The first element is the variance when the group assignment is considered as fixed, and the second element is the variance when the group assignment is considered as random, so the second element is slightly larger than the first element when with unequal allocations.
vd0
Asymptotic variance of the win difference under global null hypothesis
td
standardized wd using the variance vd
pd
2-sided p-values of td
td0
standardized wd using the variance vd0
pd0
2-sided p-value of td0
wp
Win product defined as the product of tw/tl
vp
Asymptotic variance of the win product under alterantive hypothesis
vp0
Asymptotic variance of the win product under global null hypothesis
tp
standardized log(wp) using the variancevp
pp
2-sided p-value of tp
tp0
standardized log(wp) using the variancevp0
pp0
2-sided p-value of tp0
Author(s)
Xiaodong Luo
References
Pocock S.J., Ariti C.A., Collier T. J. and Wang D. 2012. The win ratio: a new approach to the analysis of composite endpoints in clinical trials based on clinical priorities. European Heart Journal, 33, 176-182.
Luo X., Tian H., Mohanty S. and Tsai W.-Y. 2015. An alternative approach to confidence interval estimation for the win ratio statistic. Biometrics, 71, 139-145.
Bebu I. and Lachin J.M. 2016. Large sample inference for a win ratio analysis of a composite outcome based on prioritized components. Biostatistics, 17, 178-187.
See Also
Examples
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
##########################################################
## Example 1: survival (semi-competing risks) example
## with terminal event having higher priority
##########################################################
##############################
## Step 1: data generation
##############################
n<-200
rho<-0.5
b2<-0.0
b1<-0.0
bc<-1.0
lambda10<-0.1;lambda20<-0.08;lambdac0<-0.09
lam1<-rep(0,n);lam2<-rep(0,n);lamc<-rep(0,n)
z<-rep(0,n)
z[1:(n/2)]<-1
lam1<-lambda10*exp(-b1*z)
lam2<-lambda20*exp(-b2*z)
lamc<-lambdac0*exp(-bc*z)
tem<-matrix(0,ncol=3,nrow=n)
y2y<-matrix(0,nrow=n,ncol=3)
y2y[,1]<-rnorm(n);y2y[,3]<-rnorm(n)
y2y[,2]<-rho*y2y[,1]+sqrt(1-rho^2)*y2y[,3]
tem[,1]<--log(1-pnorm(y2y[,1]))/lam1
tem[,2]<--log(1-pnorm(y2y[,2]))/lam2
tem[,3]<--log(1-runif(n))/lamc
y1<-apply(tem,1,min)
y2<-apply(tem[,2:3],1,min)
d1<-as.numeric(tem[,1]<=y1)
d2<-as.numeric(tem[,2]<=y2)
###un-weighted win loss
wtest<-winratio(y1,y2,d1,d2,z)
summary(wtest)
i<-1 ##i=1,2,3,4
j<-1 ##j=1,2
###weighted win loss
wwtest<-wwratio(y1,y2,d1,d2,z,wty1=i,wty2=j)
summary(wwtest)
####general win loss
###Define the win loss function
comp<-function(y,x){
y1i<-y[1];y2i<-y[2];d1i<-y[3];d2i<-y[4]
y1j<-x[1];y2j<-x[2];d1j<-x[3];d2j<-x[4]
w2<-0;w1<-0;l2<-0;l1<-0
if (d2j==1 & y2i>=y2j) w2<-1
else if (d2i==1 & y2j>=y2i) l2<-1
if (w2==0 & l2==0 & d1j==1 & y1i>=y1j) w1<-1
else if (w2==0 & l2==0 & d1i==1 & y1j>=y1i) l1<-1
comp<-0
if (w2==1) comp<-1
else if (l2==1) comp<-(-1)
else if (w1==1) comp<-2
else if (l1==1) comp<-(-2)
comp
}
###Use the user-defined win loss function to calculate the win loss matrix
y<-cbind(y1,y2,d1,d2)
yy1<-y[z==1,]
yy0<-y[z==0,]
n1<-sum(z==1)
n0<-sum(z==0)
bindex<-matrix(0,nrow=n1,ncol=n0)
for (i in 1:n1)for (j in 1:n0)bindex[i,j]<-comp(yy1[i,],yy0[j,])
###Use the calculated win loss matrix to calculate the general win loss statistics
bgwr<-genwr(bindex)
summary(bgwr)
##################################################################
# Note: if n>=1000 or the win loss function is complex,
# one may experience long runtime. One may instead use C, C++,
# Fortran, Python to code the win loss function.
# The following provides an example using Fortran 95 code to
# define the win loss matrix and then port it back to R
# using the package "inline"
##################################################################
#####################################################
# This is to install and load package "inline"
# so that we can compile user-defined
# win loss function
#
#install.packages("inline")
library("inline")
##############################################################
# You may also need to have rtools and gcc in the PATH
# The following code add these
# for the current R session ONLY
# Please remove the '#' in the following 6 lines.
#
#rtools <- "C:\Rtools\bin"
#gcc <- "C:\Rtools\gcc-4.6.3\bin"
#path <- strsplit(Sys.getenv("PATH"), ";")[[1]]
#new_path <- c(rtools, gcc, path)
#new_path <- new_path[!duplicated(tolower(new_path))]
#Sys.setenv(PATH = paste(new_path, collapse = ";"))
##############################################################
###Define the win loss indicator by a user-supplied function
codex5 <- "
integer::i,j,indexij,d1i,d2i,d1j,d2j,w2,w1,l2,l1
double precision::y1i,y2i,y1j,y2j
do i=1,n1,1
y1i=y(i,1);y2i=y(i,2);d1i=dnint(y(i,3));d2i=dnint(y(i,4))
do j=1,n0,1
y1j=x(j,1);y2j=x(j,2);d1j=dnint(x(j,3));d2j=dnint(x(j,4))
w2=0;w1=0;l2=0;l1=0
if (d2j==1 .and. y2i>=y2j) then
w2=1
else if (d2i==1 .and. y2j>=y2i) then
l2=1
end if
if (w2==0 .and. l2==0 .and. d1j==1 .and. y1i>=y1j) then
w1=1
else if (w2==0 .and. l2==0 .and. d1i==1 .and. y1j>=y1i) then
l1=1
end if
aindex(i,j)=0
if (w2==1) then
aindex(i,j)=1
else if (l2==1) then
aindex(i,j)=-1
else if (w1==1) then
aindex(i,j)=2
else if (l1==1) then
aindex(i,j)=-2
end if
end do
end do
"
###End of defining the win loss indicator by a user-supplied function
###Convert the above code to Fortran 95 code and port it back to R
cubefnx5<-cfunction(sig = signature(n1="integer",n0="integer",p="integer",
y="numeric",x="numeric", aindex="integer"),
implicit = "none", dim = c("", "", "", "(n1,p)","(n0,p)","(n1,n0)"),
codex5, language="F95")
###Use the above defined function to calculate the win loss indicators
y<-cbind(y1,y2,d1,d2)
yy1<-y[z==1,]
yy0<-y[z==0,]
n1<-sum(z==1)
n0<-sum(z==0)
options(object.size=1.0E+10)
##The following is the win loss indicator matrix
aindex<-matrix(cubefnx5(n1,n0,length(y[1,]), yy1,yy0,
matrix(0,nrow=n1,ncol=n0))$aindex,byrow=FALSE,ncol=n0)
###Use the win loss indicator matrix to calculate the general win loss statistics
agwr<-genwr(aindex)
summary(agwr)
##########################################################
## Example 2: Continuous outcome example
## suppose there are two outcomes (y1,y2) following bivariate normal dist
## y1 is more important than y2, when comparing with (x1,x2) from another subject
## a win of first outcome if y1>x1+1 and a loss if y1<x1-1
## if tie, i.e. |y1-x1|<=1, then a win of second outcome if y2>x2+0.5
## and a loss if y2<x2-0.5. The other scenarios are tie.
##########################################################
##############################
## Step 1: data generation
##############################
n<-300
rho<-0.5
b2<-2.5
b1<-2.0
z<-rep(0,n)
z[1:(n/2)]<-1
y2y<-matrix(0,nrow=n,ncol=3)
y2y[,1]<-rnorm(n);y2y[,3]<-rnorm(n)
y2y[,2]<-rho*y2y[,1]+sqrt(1-rho^2)*y2y[,3]
y1<-b1*z+y2y[,1]
y2<-b2*z+y2y[,2]
####general win loss
###Define the win loss indicator by a user-supplied function
codex6 <- "
integer::i,j,indexij,w2,w1,l2,l1
double precision::y1i,y2i,y1j,y2j
do i=1,n1,1
y1i=y(i,1);y2i=y(i,2)
do j=1,n0,1
y1j=x(j,1);y2j=x(j,2)
w2=0;w1=0;l2=0;l1=0
if (y1i>(y1j+1.0)) then
w1=1
else if (y1i<(y1j-1.0)) then
l1=1
end if
if (w1==0 .and. l1==0 .and. y2i>(y2j+0.5)) then
w2=1
else if (w1==0 .and. l1==0 .and. y2i<(y2j-0.5)) then
l2=1
end if
aindex(i,j)=0
if (w1==1) then
aindex(i,j)=1
else if (l1==1) then
aindex(i,j)=-1
else if (w2==1) then
aindex(i,j)=2
else if (l2==1) then
aindex(i,j)=-2
end if
end do
end do
"
###End of defining the win loss indicator by a user-supplied function
###Convert the above code to Fortran 95 code and port it back to R
cubefnx6<-cfunction(sig = signature(n1="integer",n0="integer",p="integer",
y="numeric",x="numeric", aindex="integer"),
implicit = "none", dim = c("", "", "", "(n1,p)","(n0,p)","(n1,n0)"),
codex6, language="F95")
###Use the above defined function to calculate the win loss indicators
y<-cbind(y1,y2)
yy1<-y[z==1,]
yy0<-y[z==0,]
n1<-sum(z==1)
n0<-sum(z==0)
options(object.size=1.0E+10)
##The following is the win loss indicator matrix
aindex<-matrix(cubefnx6(n1,n0,length(y[1,]), yy1,yy0,
matrix(0,nrow=n1,ncol=n0))$aindex,byrow=FALSE,ncol=n0)
###Use the win loss indicator matrix to calculate the general win loss statistics
agwr<-genwr(aindex)
summary(agwr)
WWR documentation built on May 2, 2019, 11:02 a.m.
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
Description Usage Arguments Details Value Author(s) References See Also Examples
Description
Calculate the general win loss statistics and their corresponding variances under the global NULL hypothesis and under alterantive hypothesis based on Bebu and Lachin (2016) paper, which is a generalization of the win ratio of Pocock et al. (2012) and the win difference of Luo et al. (2015). This calculation needs the users to specify the win loss matrix.
Usage
1 | genwr(aindex)
|
Arguments
aindex |
a numeric matrix of win loss indicators. Suppose there are group 1 and group 0 in the study with sample sizes n_1 and n_0 respectively. The matrix |
Details
General win loss statistics
Value
n1 |
Number of subjects in group 1 |
n0 |
Number of subjects in group 0 |
n |
Total number of subjects in both groups |
totalw |
Total number of wins in group 1 |
totall |
Total number of losses in group 1 |
tw |
A vector of total numbers of wins in group 1 for each of the outcomes. Note that |
tl |
A vector of total numbers of losses in group 1 for each of the outcomes. Note that |
xp |
The ratios between |
cwindex |
The win contribution index defined as the ratio between |
clindex |
The loss contribution index defined as the ratio between |
wr |
Win ratio defined as |
vr |
Asymptotic variance of the win ratio under alterantive hypothesis |
vr0 |
Asymptotic variance of the win ratio under global null hypothesis |
tr |
standardized log(wr) using the variance |
pr |
2-sided p-value of |
tr0 |
standardized log(wr) using the variance |
pr0 |
2-sided p-value of |
wd |
Win difference defined as |
vd |
Asymptotic variance of the win difference under alterantive hypothesis. The first element is the variance when the group assignment is considered as fixed, and the second element is the variance when the group assignment is considered as random, so the second element is slightly larger than the first element when with unequal allocations. |
vd0 |
Asymptotic variance of the win difference under global null hypothesis |
td |
standardized |
pd |
2-sided p-values of |
td0 |
standardized |
pd0 |
2-sided p-value of |
wp |
Win product defined as the product of |
vp |
Asymptotic variance of the win product under alterantive hypothesis |
vp0 |
Asymptotic variance of the win product under global null hypothesis |
tp |
standardized log(wp) using the variance |
pp |
2-sided p-value of |
tp0 |
standardized log(wp) using the variance |
pp0 |
2-sided p-value of |
Author(s)
Xiaodong Luo
References
Pocock S.J., Ariti C.A., Collier T. J. and Wang D. 2012. The win ratio: a new approach to the analysis of composite endpoints in clinical trials based on clinical priorities. European Heart Journal, 33, 176-182.
Luo X., Tian H., Mohanty S. and Tsai W.-Y. 2015. An alternative approach to confidence interval estimation for the win ratio statistic. Biometrics, 71, 139-145.
Bebu I. and Lachin J.M. 2016. Large sample inference for a win ratio analysis of a composite outcome based on prioritized components. Biostatistics, 17, 178-187.
See Also
Examples
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 | ##########################################################
## Example 1: survival (semi-competing risks) example
## with terminal event having higher priority
##########################################################
##############################
## Step 1: data generation
##############################
n<-200
rho<-0.5
b2<-0.0
b1<-0.0
bc<-1.0
lambda10<-0.1;lambda20<-0.08;lambdac0<-0.09
lam1<-rep(0,n);lam2<-rep(0,n);lamc<-rep(0,n)
z<-rep(0,n)
z[1:(n/2)]<-1
lam1<-lambda10*exp(-b1*z)
lam2<-lambda20*exp(-b2*z)
lamc<-lambdac0*exp(-bc*z)
tem<-matrix(0,ncol=3,nrow=n)
y2y<-matrix(0,nrow=n,ncol=3)
y2y[,1]<-rnorm(n);y2y[,3]<-rnorm(n)
y2y[,2]<-rho*y2y[,1]+sqrt(1-rho^2)*y2y[,3]
tem[,1]<--log(1-pnorm(y2y[,1]))/lam1
tem[,2]<--log(1-pnorm(y2y[,2]))/lam2
tem[,3]<--log(1-runif(n))/lamc
y1<-apply(tem,1,min)
y2<-apply(tem[,2:3],1,min)
d1<-as.numeric(tem[,1]<=y1)
d2<-as.numeric(tem[,2]<=y2)
###un-weighted win loss
wtest<-winratio(y1,y2,d1,d2,z)
summary(wtest)
i<-1 ##i=1,2,3,4
j<-1 ##j=1,2
###weighted win loss
wwtest<-wwratio(y1,y2,d1,d2,z,wty1=i,wty2=j)
summary(wwtest)
####general win loss
###Define the win loss function
comp<-function(y,x){
y1i<-y[1];y2i<-y[2];d1i<-y[3];d2i<-y[4]
y1j<-x[1];y2j<-x[2];d1j<-x[3];d2j<-x[4]
w2<-0;w1<-0;l2<-0;l1<-0
if (d2j==1 & y2i>=y2j) w2<-1
else if (d2i==1 & y2j>=y2i) l2<-1
if (w2==0 & l2==0 & d1j==1 & y1i>=y1j) w1<-1
else if (w2==0 & l2==0 & d1i==1 & y1j>=y1i) l1<-1
comp<-0
if (w2==1) comp<-1
else if (l2==1) comp<-(-1)
else if (w1==1) comp<-2
else if (l1==1) comp<-(-2)
comp
}
###Use the user-defined win loss function to calculate the win loss matrix
y<-cbind(y1,y2,d1,d2)
yy1<-y[z==1,]
yy0<-y[z==0,]
n1<-sum(z==1)
n0<-sum(z==0)
bindex<-matrix(0,nrow=n1,ncol=n0)
for (i in 1:n1)for (j in 1:n0)bindex[i,j]<-comp(yy1[i,],yy0[j,])
###Use the calculated win loss matrix to calculate the general win loss statistics
bgwr<-genwr(bindex)
summary(bgwr)
##################################################################
# Note: if n>=1000 or the win loss function is complex,
# one may experience long runtime. One may instead use C, C++,
# Fortran, Python to code the win loss function.
# The following provides an example using Fortran 95 code to
# define the win loss matrix and then port it back to R
# using the package "inline"
##################################################################
#####################################################
# This is to install and load package "inline"
# so that we can compile user-defined
# win loss function
#
#install.packages("inline")
library("inline")
##############################################################
# You may also need to have rtools and gcc in the PATH
# The following code add these
# for the current R session ONLY
# Please remove the '#' in the following 6 lines.
#
#rtools <- "C:\Rtools\bin"
#gcc <- "C:\Rtools\gcc-4.6.3\bin"
#path <- strsplit(Sys.getenv("PATH"), ";")[[1]]
#new_path <- c(rtools, gcc, path)
#new_path <- new_path[!duplicated(tolower(new_path))]
#Sys.setenv(PATH = paste(new_path, collapse = ";"))
##############################################################
###Define the win loss indicator by a user-supplied function
codex5 <- "
integer::i,j,indexij,d1i,d2i,d1j,d2j,w2,w1,l2,l1
double precision::y1i,y2i,y1j,y2j
do i=1,n1,1
y1i=y(i,1);y2i=y(i,2);d1i=dnint(y(i,3));d2i=dnint(y(i,4))
do j=1,n0,1
y1j=x(j,1);y2j=x(j,2);d1j=dnint(x(j,3));d2j=dnint(x(j,4))
w2=0;w1=0;l2=0;l1=0
if (d2j==1 .and. y2i>=y2j) then
w2=1
else if (d2i==1 .and. y2j>=y2i) then
l2=1
end if
if (w2==0 .and. l2==0 .and. d1j==1 .and. y1i>=y1j) then
w1=1
else if (w2==0 .and. l2==0 .and. d1i==1 .and. y1j>=y1i) then
l1=1
end if
aindex(i,j)=0
if (w2==1) then
aindex(i,j)=1
else if (l2==1) then
aindex(i,j)=-1
else if (w1==1) then
aindex(i,j)=2
else if (l1==1) then
aindex(i,j)=-2
end if
end do
end do
"
###End of defining the win loss indicator by a user-supplied function
###Convert the above code to Fortran 95 code and port it back to R
cubefnx5<-cfunction(sig = signature(n1="integer",n0="integer",p="integer",
y="numeric",x="numeric", aindex="integer"),
implicit = "none", dim = c("", "", "", "(n1,p)","(n0,p)","(n1,n0)"),
codex5, language="F95")
###Use the above defined function to calculate the win loss indicators
y<-cbind(y1,y2,d1,d2)
yy1<-y[z==1,]
yy0<-y[z==0,]
n1<-sum(z==1)
n0<-sum(z==0)
options(object.size=1.0E+10)
##The following is the win loss indicator matrix
aindex<-matrix(cubefnx5(n1,n0,length(y[1,]), yy1,yy0,
matrix(0,nrow=n1,ncol=n0))$aindex,byrow=FALSE,ncol=n0)
###Use the win loss indicator matrix to calculate the general win loss statistics
agwr<-genwr(aindex)
summary(agwr)
##########################################################
## Example 2: Continuous outcome example
## suppose there are two outcomes (y1,y2) following bivariate normal dist
## y1 is more important than y2, when comparing with (x1,x2) from another subject
## a win of first outcome if y1>x1+1 and a loss if y1<x1-1
## if tie, i.e. |y1-x1|<=1, then a win of second outcome if y2>x2+0.5
## and a loss if y2<x2-0.5. The other scenarios are tie.
##########################################################
##############################
## Step 1: data generation
##############################
n<-300
rho<-0.5
b2<-2.5
b1<-2.0
z<-rep(0,n)
z[1:(n/2)]<-1
y2y<-matrix(0,nrow=n,ncol=3)
y2y[,1]<-rnorm(n);y2y[,3]<-rnorm(n)
y2y[,2]<-rho*y2y[,1]+sqrt(1-rho^2)*y2y[,3]
y1<-b1*z+y2y[,1]
y2<-b2*z+y2y[,2]
####general win loss
###Define the win loss indicator by a user-supplied function
codex6 <- "
integer::i,j,indexij,w2,w1,l2,l1
double precision::y1i,y2i,y1j,y2j
do i=1,n1,1
y1i=y(i,1);y2i=y(i,2)
do j=1,n0,1
y1j=x(j,1);y2j=x(j,2)
w2=0;w1=0;l2=0;l1=0
if (y1i>(y1j+1.0)) then
w1=1
else if (y1i<(y1j-1.0)) then
l1=1
end if
if (w1==0 .and. l1==0 .and. y2i>(y2j+0.5)) then
w2=1
else if (w1==0 .and. l1==0 .and. y2i<(y2j-0.5)) then
l2=1
end if
aindex(i,j)=0
if (w1==1) then
aindex(i,j)=1
else if (l1==1) then
aindex(i,j)=-1
else if (w2==1) then
aindex(i,j)=2
else if (l2==1) then
aindex(i,j)=-2
end if
end do
end do
"
###End of defining the win loss indicator by a user-supplied function
###Convert the above code to Fortran 95 code and port it back to R
cubefnx6<-cfunction(sig = signature(n1="integer",n0="integer",p="integer",
y="numeric",x="numeric", aindex="integer"),
implicit = "none", dim = c("", "", "", "(n1,p)","(n0,p)","(n1,n0)"),
codex6, language="F95")
###Use the above defined function to calculate the win loss indicators
y<-cbind(y1,y2)
yy1<-y[z==1,]
yy0<-y[z==0,]
n1<-sum(z==1)
n0<-sum(z==0)
options(object.size=1.0E+10)
##The following is the win loss indicator matrix
aindex<-matrix(cubefnx6(n1,n0,length(y[1,]), yy1,yy0,
matrix(0,nrow=n1,ncol=n0))$aindex,byrow=FALSE,ncol=n0)
###Use the win loss indicator matrix to calculate the general win loss statistics
agwr<-genwr(aindex)
summary(agwr)
|
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
Embedding an R snippet on your website
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.