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
|
% $Id: wapply.Rd 1441 2010-06-11 03:11:54Z warnes $
%
% $Log$
% Revision 1.10 2005/12/01 16:46:52 nj7w
% Updated Greg's email address
%
% Revision 1.9 2005/06/09 14:20:28 nj7w
% Updating the version number, and various help files to synchronize splitting of gregmisc bundle in 4 individual components.
%
% Revision 1.1.1.1 2005/05/25 22:15:31 nj7w
% Initial submission as an individual package
%
% Revision 1.8 2003/12/03 02:46:51 warnes
% - match function argument defaults with 'usage'
%
% Revision 1.7 2003/01/20 17:13:06 warnes
%
% - Updated wapply.R to allow specification of evaluation points when
% method is 'width' or 'range' using the 'pts' argument.
% - Updated wapply.Rd to add 'pts' argument
% - Fixed typos, spelling errors, gramatical errors and lack of clarity
% in wapply.Rd help text.
%
% Revision 1.6 2002/04/09 00:51:33 warneg
%
% Checkin for version 0.5.3
%
% Revision 1.5 2001/12/07 23:34:35 warneg
%
% Fixed an problem where \code was incorrectly delimited by parens
% rather than curly brackets.
%
% Revision 1.4 2001/12/05 19:36:30 warneg
% - Clarified how the width of the window is computed when method="nobs".
%
% Revision 1.3 2001/10/16 23:15:01 warneg
%
% Fixed unbalanced brace.
%
% Revision 1.2 2001/09/01 00:01:55 warneg
% Release 0.3.0
%
% Revision 1.1 2001/08/25 05:45:10 warneg
% Initial Checkin
%
%
\name{wapply}
\alias{wapply}
\title{Compute the Value of a Function Over a Local Region Of An X-Y Plot}
\description{
This function applies the specified function to the sets of y values
that are defined by overlapping "windows" in the x-dimension. For
example, setting \code{fun=mean} returns local means, while setting
\code{fun=function(x) sqrt(var(x))} returns local estimates of
the standard deviation.
}
\usage{
wapply(x, y, fun=mean, method="range", width, n=50, drop.na=TRUE,
pts, ...)
}
\arguments{
\item{x}{ vector of x values for (x,y) pairs }
\item{y}{ vector of y values for (x,y) pairs }
\item{fun}{ function to be applied }
\item{method}{ method of defining an x-neighborhood. One of
"width","nobs","range", or "fraction". See details.}
\item{width}{ width of an x-neighborhood. See details. }
\item{n}{ Number of equally spaced points at which to compute local
estimates. See details.}
\item{drop.na}{ should points which result in missing values \code{NA}
be omitted from the return value. Defaults to true. }
\item{pts}{ \code{x} locations at which to compute the local mean when
using the "width" or "range" methods. Ignored otherwise.}
\item{\dots}{ arguments to be passed to \code{fun} }
}
\details{
Two basic techniques are available for determining what points fall
within the same x-neighborhood. The first technique uses a window with
a fixed width in the x-dimension and is is selected by
setting \code{method="width"} or \code{method="range"}. For
\code{method="width"} the \code{width} argument is an absolute
distance in the x-dimension. For \code{method="range"}, the width is
expressed as a fraction of the x-range. In both cases, \code{pts}
specifies the points at which evaluation of \code{fun} occurs. When
\code{pts} is omitted, \code{n} x values equally spaced along the x
range are used.
The second technique uses windows containing k neighboring points. The
(x,y) pairs are sorted by the x-values and the nearest k/2 points with
higher x values and the k/2 nearest points with lower x values are
included in the window. When \code{method="nobs"}, k equals
\code{width} (actually 2*floor(\code{width}/2) ). When
\code{method="fraction"}, \code{width} specifies what fraction of the
total number of points should be included. The actual number of points
included in each window will be floor(n*frac/2)*2. Regardless of the
value of \code{pts}, the function \code{fun} will be evaluated at all
x locations.
}
\value{
Returns a list with components
\item{x }{x location'}
\item{y }{Result of applying \code{fun} to the window about each x location}
}
\author{ Gregory R. Warnes \email{greg@warnes.net}
}
\examples{
#show local mean and inner 2-sd interval to help diagnose changing mean
#or variance structure
x <- 1:1000
y <- rnorm(1000, mean=1, sd=1 + x/1000 )
plot(x,y)
lines(wapply(x,y,mean),col="red")
CL <- function(x,sd) mean(x)+sd*sqrt(var(x))
lines(wapply(x,y,CL,sd= 1),col="blue")
lines(wapply(x,y,CL,sd=-1),col="blue")
lines(wapply(x,y,CL,sd= 2),col="green")
lines(wapply(x,y,CL,sd=-2),col="green")
#show local mean and inner 2-sd interval to help diagnose changing mean
#or variance structure
x <- 1:1000
y <- rnorm(1000, mean=x/1000, sd=1)
plot(x,y)
lines(wapply(x,y,mean),col="red")
CL <- function(x,sd) mean(x)+sd*sqrt(var(x))
lines(wapply(x,y,CL,sd= 1,method="fraction",width=1/20),col="blue")
lines(wapply(x,y,CL,sd=-1,method="fraction",width=1/20),col="blue")
lines(wapply(x,y,CL,sd= 2,method="nobs",width=250),col="green")
lines(wapply(x,y,CL,sd=-2,method="nobs",width=250),col="green")
}
\keyword{ dplot }
|
