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
|
% Generated by roxygen2: do not edit by hand
% Please edit documentation in R/ppc-censoring.R
\name{PPC-censoring}
\alias{PPC-censoring}
\alias{ppc_km_overlay}
\alias{ppc_km_overlay_grouped}
\title{PPC censoring}
\usage{
ppc_km_overlay(
y,
yrep,
...,
status_y,
left_truncation_y = NULL,
extrapolation_factor = 1.2,
size = 0.25,
alpha = 0.7
)
ppc_km_overlay_grouped(
y,
yrep,
group,
...,
status_y,
left_truncation_y = NULL,
extrapolation_factor = 1.2,
size = 0.25,
alpha = 0.7
)
}
\arguments{
\item{y}{A vector of observations. See \strong{Details}.}
\item{yrep}{An \code{S} by \code{N} matrix of draws from the posterior (or prior)
predictive distribution. The number of rows, \code{S}, is the size of the
posterior (or prior) sample used to generate \code{yrep}. The number of columns,
\code{N} is the number of predicted observations (\code{length(y)}). The columns of
\code{yrep} should be in the same order as the data points in \code{y} for the plots
to make sense. See the \strong{Details} and \strong{Plot Descriptions} sections for
additional advice specific to particular plots.}
\item{...}{Currently only used internally.}
\item{status_y}{The status indicator for the observations from \code{y}. This must
be a numeric vector of the same length as \code{y} with values in \{0, 1\} (0 =
right censored, 1 = event).}
\item{left_truncation_y}{Optional parameter that specifies left-truncation
(delayed entry) times for the observations from \code{y}. This must be a numeric
vector of the same length as \code{y}. If \code{NULL} (default), no left-truncation
is assumed.}
\item{extrapolation_factor}{A numeric value (>=1) that controls how far the
plot is extended beyond the largest observed value in \code{y}. The default
value is 1.2, which corresponds to 20 \% extrapolation. Note that all
posterior predictive draws may not be shown by default because of the
controlled extrapolation. To display all posterior predictive draws, set
\code{extrapolation_factor = Inf}.}
\item{size, alpha}{Passed to the appropriate geom to control the appearance of
the \code{yrep} distributions.}
\item{group}{A grouping variable of the same length as \code{y}.
Will be coerced to \link[base:factor]{factor} if not already a factor.
Each value in \code{group} is interpreted as the group level pertaining
to the corresponding observation.}
}
\value{
A ggplot object that can be further customized using the \strong{ggplot2} package.
}
\description{
Compare the empirical distribution of censored data \code{y} to the
distributions of simulated/replicated data \code{yrep} from the posterior
predictive distribution. See the \strong{Plot Descriptions} section, below, for
details.
Although some of the other \pkg{bayesplot} plots can be used with censored
data, \code{ppc_km_overlay()} is currently the only plotting function designed
\emph{specifically} for censored data. We encourage you to suggest or contribute
additional plots at
\href{https://github.com/stan-dev/bayesplot}{github.com/stan-dev/bayesplot}.
}
\section{Plot Descriptions}{
\describe{
\item{\code{ppc_km_overlay()}}{
Empirical CCDF estimates of each dataset (row) in \code{yrep} are overlaid, with
the Kaplan-Meier estimate (Kaplan and Meier, 1958) for \code{y} itself on top
(and in a darker shade). This is a PPC suitable for right-censored \code{y}.
Note that the replicated data from \code{yrep} is assumed to be uncensored. Left
truncation (delayed entry) times for \code{y} can be specified using
\code{left_truncation_y}.
}
\item{\code{ppc_km_overlay_grouped()}}{
The same as \code{ppc_km_overlay()}, but with separate facets by \code{group}.
}
}
}
\examples{
\donttest{
color_scheme_set("brightblue")
# For illustrative purposes, (right-)censor values y > 110:
y <- example_y_data()
status_y <- as.numeric(y <= 110)
y <- pmin(y, 110)
# In reality, the replicated data (yrep) would be obtained from a
# model which takes the censoring of y properly into account. Here,
# for illustrative purposes, we simply use example_yrep_draws():
yrep <- example_yrep_draws()
dim(yrep)
# Overlay 25 curves
ppc_km_overlay(y, yrep[1:25, ], status_y = status_y)
# With extrapolation_factor = 1 (no extrapolation)
ppc_km_overlay(y, yrep[1:25, ], status_y = status_y, extrapolation_factor = 1)
# With extrapolation_factor = Inf (show all posterior predictive draws)
ppc_km_overlay(y, yrep[1:25, ], status_y = status_y, extrapolation_factor = Inf)
# With separate facets by group:
group <- example_group_data()
ppc_km_overlay_grouped(y, yrep[1:25, ], group = group, status_y = status_y)
# With left-truncation (delayed entry) times:
min_vals <- pmin(y, apply(yrep, 2, min))
left_truncation_y <- rep(0, length(y))
condition <- y > mean(y) / 2
left_truncation_y[condition] <- pmin(
runif(sum(condition), min = 0.6, max = 0.99) * y[condition],
min_vals[condition] - 0.001
)
ppc_km_overlay(y, yrep[1:25, ], status_y = status_y,
left_truncation_y = left_truncation_y)
}
}
\references{
Gelman, A., Carlin, J. B., Stern, H. S., Dunson, D. B., Vehtari,
A., and Rubin, D. B. (2013). \emph{Bayesian Data Analysis.} Chapman & Hall/CRC
Press, London, third edition. (Ch. 6)
Kaplan, E. L. and Meier, P. (1958). Nonparametric estimation
from incomplete observations.
\emph{Journal of the American Statistical Association}. 53(282), 457--481.
doi:10.1080/01621459.1958.10501452.
}
\seealso{
Other PPCs:
\code{\link{PPC-discrete}},
\code{\link{PPC-distributions}},
\code{\link{PPC-errors}},
\code{\link{PPC-intervals}},
\code{\link{PPC-loo}},
\code{\link{PPC-overview}},
\code{\link{PPC-scatterplots}},
\code{\link{PPC-test-statistics}}
}
\concept{PPCs}
|
