|
1
|
|
# pROC: Tools Receiver operating characteristic (ROC curves) with
|
|
2
|
|
# (partial) area under the curve, confidence intervals and comparison.
|
|
3
|
|
# Copyright (C) 2010-2014 Xavier Robin, Alexandre Hainard, Natacha Turck,
|
|
4
|
|
# Natalia Tiberti, Frédérique Lisacek, Jean-Charles Sanchez
|
|
5
|
|
# and Markus Müller
|
|
6
|
|
#
|
|
7
|
|
# This program is free software: you can redistribute it and/or modify
|
|
8
|
|
# it under the terms of the GNU General Public License as published by
|
|
9
|
|
# the Free Software Foundation, either version 3 of the License, or
|
|
10
|
|
# (at your option) any later version.
|
|
11
|
|
#
|
|
12
|
|
# This program is distributed in the hope that it will be useful,
|
|
13
|
|
# but WITHOUT ANY WARRANTY; without even the implied warranty of
|
|
14
|
|
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
|
15
|
|
# GNU General Public License for more details.
|
|
16
|
|
#
|
|
17
|
|
# You should have received a copy of the GNU General Public License
|
|
18
|
|
# along with this program. If not, see <http://www.gnu.org/licenses/>.
|
|
19
|
|
|
|
20
|
3
|
smooth <- function(...)
|
|
21
|
3
|
UseMethod("smooth")
|
|
22
|
|
|
|
23
|
|
smooth.default <- function(...) {
|
|
24
|
3
|
stats::smooth(...)
|
|
25
|
|
}
|
|
26
|
|
|
|
27
|
|
smooth.smooth.roc <- function(smooth.roc, ...) {
|
|
28
|
0
|
roc <- attr(smooth.roc, "roc")
|
|
29
|
0
|
if (is.null(roc))
|
|
30
|
0
|
stop("Cannot smooth a ROC curve generated directly with numeric 'density.controls' and 'density.cases'.")
|
|
31
|
0
|
smooth.roc(roc, ...)
|
|
32
|
|
}
|
|
33
|
|
|
|
34
|
|
smooth.roc <- function(roc, method = c("binormal", "density", "fitdistr", "logcondens", "logcondens.smooth"), n = 512, bw = "nrd0",
|
|
35
|
|
density = NULL, density.controls = density, density.cases = density,
|
|
36
|
|
start = NULL, start.controls = start, start.cases = start,
|
|
37
|
|
reuse.auc = TRUE, reuse.ci = FALSE,
|
|
38
|
|
...) {
|
|
39
|
3
|
method <- match.arg(method)
|
|
40
|
|
|
|
41
|
3
|
if (is.ordered(roc$original.predictor) && (method == "density" || method =="fitidstr"))
|
|
42
|
0
|
stop("ROC curves of ordered predictors can be smoothed only with binormal smoothing.")
|
|
43
|
|
|
|
44
|
3
|
if (method == "binormal") {
|
|
45
|
3
|
sesp <- smooth.roc.binormal(roc, n)
|
|
46
|
|
}
|
|
47
|
3
|
else if (method == "fitdistr") {
|
|
48
|
3
|
sesp <- smooth.roc.fitdistr(roc, n, density.controls, density.cases, start.controls, start.cases, ...)
|
|
49
|
|
}
|
|
50
|
3
|
else if (method == "density") {
|
|
51
|
3
|
sesp <- smooth.roc.density(roc, n, density.controls, density.cases, bw, ...)
|
|
52
|
|
}
|
|
53
|
3
|
else if (method == "logcondens") {
|
|
54
|
3
|
sesp <- smooth.roc.logcondens(roc, n)
|
|
55
|
|
}
|
|
56
|
3
|
else if (method == "logcondens.smooth") {
|
|
57
|
3
|
sesp <- smooth.roc.logcondens.smooth(roc, n)
|
|
58
|
|
}
|
|
59
|
|
else {
|
|
60
|
0
|
stop(sprintf("Impossible smooth method value '%s'. Please report this bug to <%s>.",
|
|
61
|
0
|
method, utils::packageDescription("pROC")$BugReports))
|
|
62
|
|
}
|
|
63
|
|
|
|
64
|
3
|
class(sesp) <- "smooth.roc"
|
|
65
|
3
|
sesp <- sort(sesp) # sort se and sp
|
|
66
|
|
# anchor SE/SP at 0/100
|
|
67
|
3
|
sesp$specificities <- c(0, as.vector(sesp$specificities), ifelse(roc$percent, 100, 1))
|
|
68
|
3
|
sesp$sensitivities <- c(ifelse(roc$percent, 100, 1), as.vector(sesp$sensitivities), 0)
|
|
69
|
3
|
attr(sesp, "roc") <- roc # keep the original roc. May be useful in CI.
|
|
70
|
3
|
sesp$percent <- roc$percent # keep some basic roc specifications
|
|
71
|
3
|
sesp$direction <- roc$direction
|
|
72
|
3
|
sesp$call <- match.call()
|
|
73
|
|
# keep smoothing arguments (for print and bootstrap)
|
|
74
|
3
|
sesp$smoothing.args <- list(...)
|
|
75
|
3
|
sesp$smoothing.args$method <- method
|
|
76
|
3
|
sesp$smoothing.args$n <- n
|
|
77
|
3
|
sesp$smoothing.args$start.controls <- start.controls
|
|
78
|
3
|
sesp$smoothing.args$start.cases <- start.cases
|
|
79
|
3
|
sesp$smoothing.args$density.controls <- density.controls
|
|
80
|
3
|
sesp$smoothing.args$density.cases <- density.cases
|
|
81
|
3
|
sesp$smoothing.args$bw <- bw
|
|
82
|
|
# complete fit.controls/cases if a function was passed as densfun
|
|
83
|
3
|
if (method == "fitdistr") {
|
|
84
|
3
|
if (is.null(sesp$fit.controls$densfun)) {
|
|
85
|
3
|
if (missing(density.controls))
|
|
86
|
0
|
sesp$fit.controls$densfun <- deparse(substitute(density))
|
|
87
|
|
else
|
|
88
|
3
|
sesp$fit.controls$densfun <- deparse(substitute(density.controls))
|
|
89
|
|
}
|
|
90
|
3
|
if (is.null(sesp$fit.cases$densfun)) {
|
|
91
|
3
|
if (missing(density.cases))
|
|
92
|
0
|
sesp$fit.cases$densfun <- deparse(substitute(density))
|
|
93
|
|
else
|
|
94
|
3
|
sesp$fit.cases$densfun <- deparse(substitute(density.cases))
|
|
95
|
|
}
|
|
96
|
|
}
|
|
97
|
|
|
|
98
|
|
# if there was an auc and a ci, re-do them
|
|
99
|
3
|
if (!is.null(roc$auc) && reuse.auc) {
|
|
100
|
3
|
args <- attributes(roc$auc)
|
|
101
|
3
|
args$roc <- NULL
|
|
102
|
3
|
args$smooth.roc <- sesp
|
|
103
|
3
|
sesp$auc <- do.call("auc.smooth.roc", args)
|
|
104
|
|
}
|
|
105
|
3
|
if (!is.null(roc$ci) && reuse.ci){
|
|
106
|
0
|
args <- attributes(roc$ci)
|
|
107
|
0
|
args$roc <- NULL
|
|
108
|
0
|
args$smooth.roc <- sesp
|
|
109
|
0
|
sesp$ci <- do.call(paste(class(roc$ci), "smooth.roc", sep="."), args)
|
|
110
|
|
}
|
|
111
|
|
|
|
112
|
3
|
return(sesp)
|
|
113
|
|
}
|
|
114
|
|
|
|
115
|
|
smooth.roc.density <- function(roc, n, density.controls, density.cases, bw,
|
|
116
|
|
# catch args for density
|
|
117
|
|
cut = 3, adjust = 1, kernel = window, window = "gaussian",
|
|
118
|
|
percent = roc$percent, direction = roc$direction,
|
|
119
|
|
...) {
|
|
120
|
3
|
if (!is.numeric(density.controls) || !is.numeric(density.cases)) {
|
|
121
|
3
|
predictor <- c(roc$controls, roc$cases)
|
|
122
|
3
|
if (is.character(bw)) {
|
|
123
|
3
|
bw <- match.fun(paste("bw", bw, sep="."))(predictor)
|
|
124
|
|
}
|
|
125
|
3
|
bw <- bw * adjust
|
|
126
|
3
|
from <- min(predictor) - (cut * bw)
|
|
127
|
3
|
to <- max(predictor) + (cut * bw)
|
|
128
|
|
}
|
|
129
|
3
|
if (mode(density.controls) == "function") {
|
|
130
|
3
|
density.controls <- density.controls(roc$controls, n=n, from=from, to=to, bw=bw, kernel=kernel, ...)
|
|
131
|
3
|
if (! is.numeric(density.controls)) {
|
|
132
|
0
|
if (is.list(density.controls) && !is.null(density.controls$y) && is.numeric(density.controls$y))
|
|
133
|
0
|
density.controls <- density.controls$y
|
|
134
|
|
else
|
|
135
|
0
|
stop("The 'density' function must return a numeric vector or a list with a 'y' item.")
|
|
136
|
|
}
|
|
137
|
|
}
|
|
138
|
3
|
else if (is.null(density.controls))
|
|
139
|
3
|
density.controls <- suppressWarnings(density(roc$controls, n=n, from=from, to=to, bw=bw, kernel=kernel, ...))$y
|
|
140
|
3
|
else if (! is.numeric(density.controls))
|
|
141
|
0
|
stop("'density.controls' must be either NULL, a function or numeric values of density (over the y axis).")
|
|
142
|
3
|
if (mode(density.cases) == "function") {
|
|
143
|
3
|
density.cases <- density.cases(roc$cases, n=n, from=from, to=to, bw=bw, kernel=kernel, ...)
|
|
144
|
3
|
if (! is.numeric(density.cases)) {
|
|
145
|
0
|
if (is.list(density.cases) && !is.null(density.cases$y) && is.numeric(density.cases$y))
|
|
146
|
0
|
density.cases <- density.cases$y
|
|
147
|
|
else
|
|
148
|
0
|
stop("The 'density' function must return a numeric vector or a list with a 'y' item.")
|
|
149
|
|
}
|
|
150
|
|
}
|
|
151
|
3
|
else if (is.null(density.cases))
|
|
152
|
3
|
density.cases <- suppressWarnings(density(roc$cases, n=n, from=from, to=to, bw=bw, kernel=kernel, ...))$y
|
|
153
|
3
|
else if (! is.numeric(density.cases))
|
|
154
|
0
|
stop("'density.cases' must be either NULL, a function or numeric values of density (over the y axis).")
|
|
155
|
3
|
if (length(density.controls) != length(density.cases))
|
|
156
|
0
|
stop("Length of 'density.controls' and 'density.cases' differ.")
|
|
157
|
|
|
|
158
|
3
|
perfs <- sapply((1:length(density.controls))+.5, roc.utils.perfs.dens, x=(1:length(density.controls))+.5, dens.controls=density.controls, dens.cases=density.cases, direction=direction)
|
|
159
|
|
|
|
160
|
3
|
return(list(sensitivities = perfs[2,] * ifelse(percent, 100, 1),
|
|
161
|
3
|
specificities = perfs[1,] * ifelse(percent, 100, 1)))
|
|
162
|
|
}
|
|
163
|
|
|
|
164
|
|
smooth.roc.binormal <- function(roc, n) {
|
|
165
|
3
|
df <- data.frame(sp=qnorm(roc$sp * ifelse(roc$percent, 1/100, 1)), se=qnorm(roc$se * ifelse(roc$percent, 1/100, 1)))
|
|
166
|
3
|
df <- df[apply(df, 1, function(x) all(is.finite(x))),]
|
|
167
|
3
|
if (dim(df)[1] <= 1) # ROC curve or with only 1 point
|
|
168
|
0
|
stop("ROC curve not smoothable (not enough points).")
|
|
169
|
3
|
model <- lm(sp~se, df)
|
|
170
|
3
|
if(any(is.na(model$coefficients[2])))
|
|
171
|
0
|
stop("ROC curve not smoothable (not enough points).")
|
|
172
|
3
|
se <- qnorm(seq(0, 1, 1/(n-1)))
|
|
173
|
3
|
sp <- predict(model, data.frame(se))
|
|
174
|
|
|
|
175
|
3
|
return(list(sensitivities = pnorm(se) * ifelse(roc$percent, 100, 1),
|
|
176
|
3
|
specificities = pnorm(sp) * ifelse(roc$percent, 100, 1),
|
|
177
|
3
|
model = model))
|
|
178
|
|
}
|
|
179
|
|
|
|
180
|
|
smooth.roc.fitdistr <- function(roc, n, densfun.controls, densfun.cases, start.controls, start.cases, ...) {
|
|
181
|
3
|
load.suggested.package("MASS")
|
|
182
|
|
|
|
183
|
3
|
densfuns.list <- list(beta = "dbeta", cauchy = "dcauchy", "chi-squared" = "dchisq", exponential = "dexp", f = "df",
|
|
184
|
3
|
gamma = "dgamma", geometric = "dgeom", "log-normal" = "dlnorm", lognormal = "dlnorm",
|
|
185
|
3
|
logistic = "dlogis", "negative binomial" = "dnbinom", normal = "dnorm", poisson = "dpois",
|
|
186
|
3
|
t = "dt", weibull = "dweibull")
|
|
187
|
|
|
|
188
|
3
|
if (is.null(densfun.controls))
|
|
189
|
3
|
densfun.controls <- "normal"
|
|
190
|
3
|
else if (is.character(densfun.controls))
|
|
191
|
3
|
densfun.controls <- match.arg(densfun.controls, names(densfuns.list))
|
|
192
|
|
|
|
193
|
3
|
if (is.null(densfun.cases))
|
|
194
|
3
|
densfun.cases <- "normal"
|
|
195
|
3
|
else if (is.character(densfun.cases))
|
|
196
|
3
|
densfun.cases <- match.arg(densfun.cases, names(densfuns.list))
|
|
197
|
|
|
|
198
|
3
|
fit.controls <- MASS::fitdistr(roc$controls, densfun.controls, start.controls, ...)
|
|
199
|
3
|
fit.cases <- MASS::fitdistr(roc$cases, densfun.cases, start.cases, ...)
|
|
200
|
|
|
|
201
|
|
# store function name in fitting results
|
|
202
|
3
|
if (mode(densfun.controls) != "function")
|
|
203
|
3
|
fit.controls$densfun <- densfun.controls
|
|
204
|
3
|
if (mode(densfun.cases) != "function")
|
|
205
|
3
|
fit.cases$densfun <- densfun.cases
|
|
206
|
|
|
|
207
|
3
|
x <- seq(min(c(roc$controls, roc$cases)), max(c(roc$controls, roc$cases)), length.out=n)
|
|
208
|
|
|
|
209
|
|
# get the actual function name for do.call
|
|
210
|
3
|
if (is.character(densfun.controls))
|
|
211
|
3
|
densfun.controls <- match.fun(densfuns.list[[densfun.controls]])
|
|
212
|
3
|
if (is.character(densfun.cases))
|
|
213
|
3
|
densfun.cases <- match.fun(densfuns.list[[densfun.cases]])
|
|
214
|
|
|
|
215
|
|
# ... that should be passed to densfun
|
|
216
|
3
|
if (length(list(...)) > 0 && any(names(list(...)) %in% names(formals(densfun.controls))))
|
|
217
|
0
|
dots.controls <- list(...)[names(formals(densfun.controls))[match(names(list(...)), names(formals(densfun.controls)))]]
|
|
218
|
|
else
|
|
219
|
3
|
dots.controls <- list()
|
|
220
|
3
|
if (length(list(...)) > 0 && any(names(list(...)) %in% names(formals(densfun.cases))))
|
|
221
|
0
|
dots.cases <- list(...)[names(formals(densfun.cases))[match(names(list(...)), names(formals(densfun.cases)))]]
|
|
222
|
|
else
|
|
223
|
3
|
dots.cases <- list()
|
|
224
|
|
|
|
225
|
3
|
density.controls <- do.call(densfun.controls, c(list(x=x), fit.controls$estimate, dots.controls))
|
|
226
|
3
|
density.cases <- do.call(densfun.cases, c(list(x=x), fit.cases$estimate, dots.cases))
|
|
227
|
|
|
|
228
|
3
|
perfs <- sapply(x, roc.utils.perfs.dens, x=x, dens.controls=density.controls, dens.cases=density.cases, direction=roc$direction)
|
|
229
|
|
|
|
230
|
3
|
return(list(sensitivities = perfs[2,] * ifelse(roc$percent, 100, 1),
|
|
231
|
3
|
specificities = perfs[1,] * ifelse(roc$percent, 100, 1),
|
|
232
|
3
|
fit.controls=fit.controls, fit.cases=fit.cases))
|
|
233
|
|
}
|
|
234
|
|
|
|
235
|
|
smooth.roc.logcondens <- function(roc, n) {
|
|
236
|
3
|
load.suggested.package("logcondens")
|
|
237
|
|
|
|
238
|
3
|
sp <- seq(0, 1, 1/(n-1))
|
|
239
|
3
|
logcondens <- logcondens::logConROC(roc$cases, roc$controls, sp)
|
|
240
|
3
|
se <- logcondens$fROC
|
|
241
|
|
|
|
242
|
3
|
return(list(sensitivities = se * ifelse(roc$percent, 100, 1),
|
|
243
|
3
|
specificities = (1 - sp) * ifelse(roc$percent, 100, 1),
|
|
244
|
3
|
logcondens = logcondens))
|
|
245
|
|
}
|
|
246
|
|
|
|
247
|
|
smooth.roc.logcondens.smooth <- function(roc, n) {
|
|
248
|
3
|
load.suggested.package("logcondens")
|
|
249
|
|
|
|
250
|
3
|
sp <- seq(0, 1, 1/(n-1))
|
|
251
|
3
|
logcondens <- logcondens::logConROC(roc$cases, roc$controls, sp)
|
|
252
|
3
|
se <- logcondens$fROC.smooth
|
|
253
|
|
|
|
254
|
3
|
return(list(sensitivities = se * ifelse(roc$percent, 100, 1),
|
|
255
|
3
|
specificities = (1 - sp) * ifelse(roc$percent, 100, 1),
|
|
256
|
3
|
logcondens = logcondens))
|
|
257
|
|
}
|
xrobin