Sliderfns

Some Slider Functions
Hans Peter Wolf
pw090615-1650, /home/wiwi/pwolf/R/aplpack/sliderfns, File: sliderfns.rev
March 13, 2012
Contents
1 Slider functions for exploratory data analysis 3
2 Slider functions – overview and examples 4
2.1 The functions of the paper . . . . . . . . . . . . . . . . . . . . . . 4
2.2 slider.hist . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.3 slider.density . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.4 slider.brush.pairs . . . . . . . . . . . . . . . . . . . . . . . . 7
2.5 slider.brush.plot.xy . . . . . . . . . . . . . . . . . . . . . . . 8
2.6 slider.split.plot.ts . . . . . . . . . . . . . . . . . . . . . . . 9
2.7 slider.zoom.plot.ts . . . . . . . . . . . . . . . . . . . . . . . . 10
2.8 slider.smooth.plot.ts . . . . . . . . . . . . . . . . . . . . . . 11
2.9 slider.lowess.plot . . . . . . . . . . . . . . . . . . . . . . . . 12
2.10 slider.bootstrap.lm.plot . . . . . . . . . . . . . . . . . . . . 13
3 General issues of the implementation 14
3.1 The structure of slider functions . . . . . . . . . . . . . . . . . . 14
3.1.1 A very simple example . . . . . . . . . . . . . . . . . . . . 14
3.1.2 The main arguments of slider . . . . . . . . . . . . . . . 15
3.1.3 Buttons and slider variables . . . . . . . . . . . . . . . . . 16
3.2 General Rules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
3.2.1 Names of the slider functions . . . . . . . . . . . . . . . . 18
3.2.2 Formal arguments for data sets . . . . . . . . . . . . . . . 18
3.2.3 Passing additional arguments . . . . . . . . . . . . . . . . 20
3.3 Density Presenter . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
4 Histograms and Density Traces 24
4.1 Class number of histograms – slider.hist . . . . . . . . . . . . 24
4.1.1 Test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
1

4.2 Width and kernel of a density trace – slider.density . . . . . . 25
4.2.1 Test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
4.2.2 Help Page . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
5 Brushing Functions 28
5.1 A draftsman’s display with Brushing – slider.brush.pairs . . 28
5.1.1 Test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
5.2 Scatter plot brushing – slider.brush.plot.xy . . . . . . . . . . 29
5.2.1 Test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
5.2.2 Help Page . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
6 Time Series Plots 31
6.1 Splitted time series – slider.split.plot.ts . . . . . . . . . . . 31
6.1.1 Test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
6.1.2 Help Page . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
6.2 Zooming in Time Series – slider.zoom.plot.ts . . . . . . . . . 34
6.2.1 Test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
6.2.2 Help Page . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
6.3 Smoothing Time Series – slider.smooth.plot.ts . . . . . . . . 36
6.3.1 Test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
6.3.2 Help Page . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
7 Regression 38
7.1 Smoothing by lowess – slider.lowess.plot . . . . . . . . . . . 38
7.1.1 Test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
7.1.2 Help Page . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
7.2 Sensitivity of Regression Line – slider.bootstrap.lm.plot . . 39
7.2.1 Test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
7.2.2 Help Page . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
8 Appendix 41
8.1 Deﬁnition of slider . . . . . . . . . . . . . . . . . . . . . . . . . 41
2

1 Slider functions for exploratory data analysis
In this paper we discuss a set of slider functions for constructing dynamical sta-
tistical plots. Some of the characteristics of these plots are controlled by sliders
and you can modify the plots by the sliders. The sliders will be generated within
Tcl/Tk widgets. Moving a slider results in changing the associated parameter
and the graphics output is updated. For example, slider.density() plots a
density trace and opens a Tcl/Tk widget with one slider that is linked to the
width parameter of density(). After shifting the slider and releasing the mouse
button the density trace is plotted again taking into account the new window
width.
The next section contains a list of functions presented in this paper. Then you
get some examples so you can see the function in action. Some special tricks are
used to customize the behaviour of the slider functions. These programming
pearls are discussed in section three. In the remaining sections you find the
definitions of the slider function. We hope that the functions are useful for you
in the way they are defined. However, you are invited to take a closer look at the
code. Sometimes it will be very easy to enhance the proposals and you quickly
get new slider functions.
3

2 Slider functions – overview and examples
2.1 The functions of the paper
The following list shows the functions available in this package:
1 linux command to ﬁnd the slider functions 1 ≡
fns<-readLines("sliderfns.rev")
fns<-fns[grep("^slider.*function",fns)]
sort(unique(sub("<-.*$","",fns)))
Wed Mar 25 10:12:18 2009
[1] "slider" "slider.bootstrap.lm.plot"
[3] "slider.brush.pairs" "slider.brush.plot.xy"
[5] "slider.density" "slider.hist"
[7] "slider.lowess.plot" "slider.present.density"
[9] "slider.show.density" "slider.show.norm"
[11] "slider.show.normal.density" "slider.smooth.plot.ts"
[13] "slider.split.plot.ts" "slider.trimmed.mean"
[15] "slider.trimmed.mean.plus" "slider.trimmed.mean.simple"
[17] "slider.xyz" "slider.zoom.plot.ts"
4

The next subsections contain a sequence of screen shots showing the some of
the slider functions in action.
2.2 slider.hist
slider.hist computes histograms. The number of classes is controlled by a
slider. As a special feature slider.hist has a panel argument. This argument
allows you to deliver a panel function which draws some additional plotting
elements. In the following example panel is used to add a normal density curve
as well as a rug.
2 show slider.hist 2 ≡
deﬁne slider.hist 31
slider.hist(rivers,xlab="x",col="green",probability=TRUE,
panel=function(x){
xx<-seq(min(x),max(x),length=100)
yy<-dnorm(xx,mean(x),sd(x))
lines(xx,yy); rug(x); print(summary(yy))
}
)
After moving the slider to position 17 the control widget looks like this:
After releasing the mouse button the graphics device shows the following plot:
You see that the ﬁt of the normal distribution is not very good. Perhaps an
exponential distribution is a quite better model. Can you build a new call of
slider.hist with a suitable exponential density curve? Hint: You have to
exchange the line of code that computes the normal density values:
remove yy<-dnorm(xx,mean(x),sd(x))
write yy<-dexp(xx,1/mean(x)).
5

2.3 slider.density
slider.density computes density traces. The parameter width and the type
of the kernel function are controlled by sliders. Like slider.hist the function
slider.density has a panel argument to add further graphical elements. In
the example slider.density a normal density curve and a rug are added by
the panel function.
3 show slider.density 3 ≡
deﬁne slider.density 34
slider.density(log(rivers),xlab="rivers",col="red",
panel=function(x){
}
)
If you want to select a width of 3 percent of the range of the data and a rectan-
gular kernel the slider control widget must have the following apearance. With
these settings the graph will be very shaky.
The parameter settings "width=20", "kernel=epanechnikov" result in a much
smoother plot.
6

2.4 slider.brush.pairs
pairs() constructs a draftsman’s display – a set of two dimensional scatter-
plots. This function is really useful for data sets with 3 to 10 variables because
dependencies between a pair of variables can be seen at once.
However, sometimes it is very interesting to explore two dimensional dependen-
cies of a subset of the data points that are defined by a third variable. This idea
leads to the function slider.brush.pairs which computes a pairs plot and
allows you to define an interval for the values of one variable. After changing
the settings all data points of the visible interval are recolored using the color
”red”.
4 show slider.brush.pairs 4 ≡
define slider.brush.pairs 38
slider.brush.pairs(iris[,-4])
7

2.5 slider.brush.plot.xy
The function slider.brush.plot.xy is a small version of slider.brush.pairs().
It is suitable for three dimensional data sets. slider.brush.plot.xy() com-
putes a two dimensional scatterplot and chooses the color of the points con-
ditional to a third variable. Therefore, you can start the exploration with
slider.brush.pairs() and if you want to concentrate on one of the scatter-
plots — e. g. for the purpose of presentation — you can call slider.brush.plot.xy().
5 show of slider.brush.plot.xy 5 ≡
deﬁne slider.brush.plot.xy 40
slider.brush.plot.xy(iris[,1:3])
8

2.6 slider.split.plot.ts
Time series often have different structures in the distinct intervals. Then it may
be interesting to inspect various subsets of the data. For comparing different
intervals (in time) you can split the whole time series into pieces and fit linear
regression models to the various regions. The function slider.split.plot.ts
has been designed for doing this job. By sliders you are able to define a point on
the time axis as well as the length of a time interval (strip). Then a time series
plot of your data is computed and it is partitioned into stripes in a way that
one limit between two strip is identical with the selected point. Furthermore,
regression lines of fitted linear models are drawn within each of the intervals.
Alternatively slider.split.plot.ts allows you to draw the main characteris-
tics of the subsets of points. A mouse click on the button fivenum summary of
the control widget will add graphical five number summaries of the data values
of each region.
6 show slider.split.plot.ts 6 ≡
define slider.split.plot.ts 43
slider.split.plot.ts(as.vector(sunspots)[1:100])
9

2.7 slider.zoom.plot.ts
Some time series are often very long. For inspection such a series you want to
be able to view the data of selected intervals. slider.zoom.plot.ts enables
you to choose two windows of the whole range of the data and the function
computes two time series plots of the two sections. With this function it is easy
to compare the structures of two diﬀerent regions of a time series.
7 show slider.zoom.plot.ts 7 ≡
deﬁne slider.zoom.plot.ts 46
slider.zoom.plot.ts(as.vector(sunspots),2)
10

2.8 slider.smooth.plot.ts
The function slider.smooth.plot.ts allows you to smooth a time series by
running medians or some other operations which has been proposed by Tukey.
The implementation is based on smooth() and the operation is applied by click-
ing one of the buttons of the control widget.
8 show slider.smooth.plot.ts 8 ≡
deﬁne slider.smooth.plot.ts 50
slider.smooth.plot.ts(rnorm(100))
11

2.9 slider.lowess.plot
lowess() is an R function for analyzing the trend of points in a scatterplot.
The smoothness of the computed line depends on a span parameter. Because
of the algorithm is time consuming lowess has a parameter to ﬁx the num-
ber of iterations to be done. The function slider.lowess.plot() computes a
scatterplot of a data set and lets you select these two parameters by an control
widget.
9 show slider.lowess.plot 9 ≡
deﬁne slider.lowess.plot 53
slider.lowess.plot(cars)
12

2.10 slider.bootstrap.lm.plot
Sometimes you have fitted a model curve on the base of a data set and you
wonder how sensitive the result is against changes of the data. The function
slider.bootstrap.lm.plot() draws a scatterplot containing the curve of a
fitted model as well as some curves that have been computed on samples of the
original data points. The user is allowed to select the degree of the polynom to
be fitted, the number of bootstrap samples and the random seed.
10 show slider.bootstrap.lm.plot 10 ≡
define slider.bootstrap.lm.plot 56
daten<-iris[,2:3]
slider.bootstrap.lm.plot(daten)
13

3 General issues of the implementation
3.1 The structure of slider functions
The last section discusses some slider functions and the reader has got examples
of their application. Now we will explore the structure of the slider functions. All
of them call the function slider to create and to control Tcl/TK widgets. The
widgets consist of sliders and buttons that are linked to R functions describing
the actions to be performed.
To demonstrate the usage of slider within the slider functions we discuss some
examples now. As a very simple one let us build a slider function that prints the
trimmed mean of a data set depending on the trim fraction which is controlled
by a slider.
3.1.1 A very simple example
In this paragraph we define the function slider.trimmed.mean.simple. It
shows the usage of the function slider() for doing the jobs
• create a control widget with one slider
• define the effect or the action to be done in case of a slider movement
• link the slider with the action.
Before going into details activate the following code chunk, please. It will define
the function "slider.trimmed.mean.simple" and it will call the new function
with the data set "rivers[1:50]":
11 define slider.trimmed.mean 11 ≡
slider.trimmed.mean.simple<-function(x)
{
# general initialisations
if(missing(x)) return("Error: no data found!")
# function to update result
compute.trimmed.mean<-function(...){
# get slider
alpha <- slider(no=1)
# print result
cat(paste("trim fraction",alpha,":",mean(x,trim=alpha),"n"))
}
# definition of slider widget
slider(compute.trimmed.mean,"trim fraction",0,0.5,0.01,0)
# initial computation
compute.trimmed.mean()
}
slider.trimmed.mean.simple(rivers[1:50])
Move the slider named "trim fraction" and observe what happens! Now let’s
take a look at the definition of slider.trimmed.mean.simple. The function
consists of four instructions:
• The first statement is an input check – business as usual.
14

• The second one defines the effect of a movement of the slider in form of
an action function although the slider hasn’t been defined yet.
• Then the function slider is called to build the control widget and the
action function is passed by the first parameter.
• At last the action function is called and a trimmed mean will be printed.
You see slider.trimmed.mean.simple() mainly consists of the definition of
the action function compute.trimmed.mean() and a call of slider().
compute.trimmed.mean() fulfills two tasks: fetching the actual value of the
slider variable (trim fraction alpha) and printing the trimmed mean. slider(no=1)
reads the value of slider number "no=1" and then the value is assigned to the
local variable alpha. The computation of the trimmed mean by mean() and its
printing follows.
3.1.2 The main arguments of slider
A slider of a slider control widget is declared by six parameters. The meaning
of them is found in following list.
1. sl.functions: an effect function which is called after moving the slider
2. sl.names: a text for labeling the slider
3. sl.mins: the minimum of the range of the slider
4. sl.maxs: the maximum of the range of the slider
5. sl.deltas: the increment of the scale of the slider
6. sl.defaults: the default value of the slider
These parameters are arguments of the function slider. The following exam-
ple has the same structure as slider.trimmed.mean.simple(). However, the
action function computes a plot and in the call of slider() the names of the
formal parameters are used.
12 define slider.trimmed.mean 11 + ≡
slider.trimmed.mean<-function(x)
{
# function to update plot
refresh<-function(...){
# get slider
# compute mean
tm <- mean(x,trim=alpha); m <- mean(x)
# construct plot
plot(as.data.frame(x))
abline(v=tm,col="red")
text(tm,1.1,adj=1,"trimmed mean",col="red")
abline(v=m,col="black")
text( m,0.9,adj=0,"mean")
title(paste("trimmed mean:", format(tm,signif=4),
15

"ntrim fraction:",round(alpha*100),"%"))
}
slider(sl.functions=refresh,
sl.names= "trim fraction",
sl.mins= 0,
sl.maxs= 0.5,
sl.deltas= 0.01,
sl.defaults= 0
)
# initial plot
refresh()
}
slider.trimmed.mean(rivers[1:50])
It may be astonishing that the argument names are in plural form. This indicates
that more than one slider can be defined in the same way. The following section
will show you an example with two sliders.
3.1.3 Buttons and slider variables
The next version of the running example — slider.trimmed.mean.plus() —
demonstrates how two sliders are implemented. To keeps things simple a second
slider is defined. It allows the user to change the color of the plot. In addition
three buttons are generated to switch between three graphical representations
of the data: dotplot, histogram or boxplot. The code is a little bit longer.
However, we are shure that it isn’t very difficult to understand it.
13 define slider.trimmed.mean 11 + ≡
slider.trimmed.mean.plus<-function(x)
{
slider(obj.name="ptype",obj.value="hist")
# function to update plot
# get settings
16

mycol <- slider(no=2)
ptype <- slider(obj.name="ptype")
# computation
tm <- mean(x,trim=alpha); m <- mean(x)
# construct high level plot
if(ptype=="dotplot") plot(as.data.frame(x),col=mycol)
if(ptype=="hist") hist(x,col=mycol)
if(ptype=="boxplot") boxplot(x,col=mycol,horizontal=TRUE)
# perform some lower level jobs
abline(v=tm,col="red")
text(tm,1.1,adj=1,"trimmed mean",col="red")
abline(v=m,col="black")
text( m,0.9,adj=0,"mean")
title(paste("trimmed mean:", format(tm,signif=4),
"ntrim fraction:",round(alpha*100),"%"))
}
# three functions to react on buttons
b1<-function(...){slider(obj.name="ptype",obj.value="dotplot");refresh()}
b2<-function(...){slider(obj.name="ptype",obj.value="hist"); refresh()}
b3<-function(...){slider(obj.name="ptype",obj.value="boxplot");refresh()}
slider(sl.functions= refresh,
sl.names= c("trim fraction","color"),
sl.mins= c(0,1),
sl.maxs= c(0.5,13),
sl.deltas= c(0.01,1),
sl.defaults= c(0,1),
but.functions=c(b1,b2,b3),
but.names= c("dotplot","histogram","boxplot")
)
# initial plot
refresh()
}
slider.trimmed.mean.plus(rivers[1:50])
17

In this example vectors of length two are assigned to the arguments sl.names,
sl.mins, sl.maxs, sl.deltas, and sl.defaults. The first elements of the
vectors describe the settings of the first slider whereas the second ones fix the
parameters of the other slider.
The buttons are implemented by a vector of threee names (but.names) and by a
vector of three effect functions (but.functions). To handle the state of the se-
lected plot type a slider variable (ptype) is created. During updating the plot the
actual value of ptype is found by the statement slider(obj.name="ptype").
After changing ptype it is stored by statements like slider(obj.name="ptype",
obj.value="dotplot"). In this example the effect functions of the buttons only
consist of a statement to store the plot type and to update the plot by calling
the effect function refresh().
In summary there is a simple structure for implementing sliders and buttons.
Slider variables can be used to communicate states, values and other informa-
tion.
3.2 General Rules
In the last section the reader learned the syntax of the function slider and he
will be able to write slider function of his own. To achieve a similar structure
of the slider functions introduced in this paper we propose some general rules
now.
3.2.1 Names of the slider functions
The names of the slider functions of this paper always start with the character
string "slider" and end with a string indicating the job to be done, e.g. a
statistical procedure or function like hist, pairs or smooth.
3.2.2 Formal arguments for data sets
One dimensional data sets are passed to a slider* function by the argument x.
If the data set is of two dimensions there are two types of calls: Either argument
x is a matrix or a data frame with two columns or the argruments x and y are
used for the first and the second variable. If x is a vector and y not available
then x is used as the second dimension and seq(x) as the first one. Three
dimensional data sets are managed in an analogous way: A (n × 3)-matrix has
to be assigned to x or the arguments x, y and z must be used. These rules
enable us to write some general statements for checking and for preparation of
some local variables containing the data.
Checking the one dimensional case is simple. Besides x.name stores how the x
argument has been assigned by the user.
14 check if x is a vector 14 ≡ ⊂ 31, 34, 38, 66
x.name<-deparse(substitute(x))
18

if(missing(x)||length(x)<2) return("Error: x must be a vector")
To check a two dimensional data input the two cases described above has to be
handled. After checking the data are assigned to two local variables x and y.
15 check if x is a matrix or x and y are vectors 15 ≡ ⊂ 53, 56
y.name<-deparse(substitute(y))
if(length(x)<2) return("Error: x is of length 0 or 1")
if(!is.null(y)){
if(length(y)<2) return("Error: y must be a vector")
if(length(x)!=length(y))
return("Error: x and y must have the same length")
x<-cbind(x,y)
}
if(!is.matrix(x) && !is.data.frame(x)){
x<-cbind(seq(x),x)
y.name<-x.name; x.name<-"index"
}
if(is.null(y.name)){x.name<-colnames(x)[1]; y.name<-colnames(x)[2]}
y<-x[,2]; x<-x[,1]
In the case of three dimensions the arguments y and z are dealt like y in the
two dimensional case.
16 check if x, y,z are vectors or x is matrix with >2 columns 16 ≡ ⊂ 40
y.name<-deparse(substitute(y))
z.name<-deparse(substitute(z))
if(length(x)<2) return("Error: x is of length 0 or 1")
if(!is.null(y)){
if(length(y)<2) return("Error: y must be a vector")
if(length(x)!=length(y))
return("Error: x and y must have the same length")
x<-cbind(x,y)
}
if(!is.null(z)){
if(length(z)<2) return("Error: z must be a vector")
if(length(x)!=length(z))
return("Error: x and z must have the same length")
x<-cbind(x,z)
}
if(!is.matrix(x)&& !is.data.frame(x) && ncol(x)<3)
return("Error: not enough variables")
if("NULL"==y.name){x.name<-colnames(x)[1]; y.name<-colnames(x)[2]}
if("NULL"==z.name){x.name<-colnames(x)[1]; z.name<-colnames(x)[3]}
z<-x[,3]; y<-x[,2]; x<-x[,1]
Time series are diﬀerent from a one dimensional data set because data points
are connected with times. Therefore, we need a code chunk for checking time
series.
17 check if x is a time series 17 ≡ ⊂ 43, 46, 50
if(missing(x)||length(x)<2) return("Error: x must be a vector")
y<-x; if(is.ts(x)) { x<-time(x) } else { x<-seq(x) }
19

3.2.3 Passing additional arguments
Remark: This section should be skipped during the first reading of the text!
You know that high level plotting function have the formal parameter "...".
This argument allows the user to pass additional settings to the graphics device.
At once the question arises: Can we manage to pass such arguments to plotting
functions that are called by a slider function? For example: a histogram is
computed in the refresh function and the user wants to produce green bars.
There are different answers to tackle this problem. At first you can consider
which of the graphics parameters are relevant to be set by the user. Then we
can introduce these parameter in the header of a slider function. But the list of
parameters will be very long because the user may be want to modify attributes
of scale, point size, type, color, etc.
A second way is to define a "..." argument for passing additional parameters.
slider.xyz<-function(x,...)
But there will be some difficulties to be managed if you use the "..." construc-
tion. How do we handle the "..." argument, how do we set default values for
some parameter and how can we remove parameters that must not be changed
at all? In the initialization part of the slider.* functions it is possible to mod-
ify the list of the additional arguments. To demonstrate this step we show how
you can remove a breaks argument and then append a "main" argument. If
main is not available it is set to x.name . (Cleary there is no graphics parameter
with name breaks but users used to set breaks in their calls of hist may try to
deliver a breaks argument.)
18 modification of the "..." argument, not for evaluation 18 ≡
args<-list(...)
args<-args[names(args)!="breaks"] # remove "breaks"
if(!any("main"==names(args))) # set default "main"
args<-c(args,list(main=x.name))
Now the argument list args has to be passed to the function hist within the
action or refresh function. Writing hist(x,args) doesn’t work because the
second argument of hist is the break argument and not a slot for further
arguments.
The second idea will be to use the "..." of hist: hist(x,...=args). But
the "..." argument of hist is passed to the functions title and axis and
so we don’t get green bars at all. Our answer to this problem is to describe
the call by its elements: the function hist should be called with the argument
list consisting of the additional arguments stored in args and other arguments
specifying the data set and the breaks, for example. Then the function do.call
is called with theses elements. Keep in mind that do.call needs the arguments
as an alist.
To summarize the following construction will work and is used in the function
slider.hist:
19 some details of a slider function, not for evaluation 19 ≡
slider.xyz<-function(x,...)
{
initialisation statements NA
args<-list(...)
20

args<-args[names(args)!="breaks"] # remove "breaks"
if(!any("main"==names(args))) # set default "main"
args<-c(args,list(main=x.name))
more initialisation statements NA
some computations for the refresh task NA
do.call("hist",c(alist(x=x,breaks=breaks),args))
further refreshing statements NA
}
tail of function slider.xyz NA
}
Before we close this section we will present a further proposal to deal with the
arguments. In the game there are four sets of information:
• the set of default settings coded in the slider function
• the set of additional parameters deﬁned by the user
• the set of forbidden arguments that must not be used
• the set of important settings that must not be changed by the user
We can save these sets on suitable list objects:
20 * 20 ≡
# defaults (these are allowed to be changed by users)
defaults<-list(main=x.name)
# graphics parameter of the user
args<-list(...)
# forbidden (these must not be passed to a slider function)
set.of.forbidden.args<-list(probability=TRUE)
# important (these are passed and must not to be changed by users)
important<-alist(bty="n",sub="graphics computed by slider")
Now the sets have to be combined. At ﬁrst we remove the forbidden elements
from the argument list of the user. Then we concatenate the important ones,
the user arguments and the default settings. Finally, we remove elements whose
names occur on the list earlier.
21 * 20 + ≡
# remove not allowed
args<-args[!names(args) %in% set.of.forbidden.args]
# set of all parameters
defaults<-c(important,args,defaults)
# removing of double ones
args<-defaults[unique(names(defaults))]
The refresh function has a very simple structure now.
22 * 20 + ≡
xrange<-range(x); num<-slider(no=1)
breaks<-seq(xrange[1],xrange[2],length=num+1)
do.call("hist",c(alist(x=x,breaks=breaks),args ))
panel(x)
}
21

In the following we have not implemented these ideas completely. Sometimes
we pass ... arguments without further checks.
3.3 Density Presenter
In this paragraph we discuss some tools for plotting densities of distributions.
The parameters of the distribution are controlled by sliders. These functions
allow the user to study how the shapes of the densities depend on the parameter
settings. It is a very easy exercise to find an implementation for a special
distribution, e.g., the normal distribution.
23 define slider.show.normal.density 23 ≡ ⊂ 24, 30
slider.show.normal.density<-function(name,...)
{
x<-seq(-10,10,length=200)
args<-list(...)
par1<-slider(no=1); par2<-slider(no=2)
f.x<-dnorm(x,par1,par2)
main=paste("normal distributionn",
"E(X) =",format(par1,digits=3),
", sd(X) =",format(par2,digits=3))
# plot(x,f.x,type="l",main=main)
do.call("plot",c(alist(x=x,y=f.x,type="l",main=main),args))
}
slider(refresh,
c("expectation","standard deviation"),
c(-20,.001),c(20,20),c(.1,.1),c(0.1,1)
)
refresh()
cat("use slider to select parameters!n")
}
Here comes a check.
24 * 20 + ≡
define slider.show.normal.density 23
slider.show.normal.density(col="red",bty="n")
A more interesting idea is to write a function that operates as a general presen-
ter. The function slider.show.density() shows the shapes of the densities of
various distributions.
After the user has collected some experiences she or he will like to write a
improved version. Before you start programming your own presenter analyze
the following function:
25 define slider.show.density 25 ≡ ⊂ 30
slider.show.density<-function(
distribution="norm",
mins=c(-100,-100,-50,.001),
maxs=c(100,100,50,100),
deltas=c(1,1,.1,.1),
defaults=c(-5,5,0,1),type="l",...)
{
args<-list(...)
lim1<-slider(no=1); lim2<-slider(no=2)
22

par1<-slider(no=3); par2<-slider(no=4)
x<-seq(min(lim1,lim2),max(lim1,lim2),length=200)
ddist<-paste("d",distribution,sep="")
f.x<-do.call(ddist,alist(x=x,par1,par2))
main=paste("distribution:",distribution,"n",
"parameter 1: ",format(par1,digits=3),
", parameter 2: =",format(par2,digits=3))
# plot(x,f.x,type=type,main=main)
do.call("plot",c(alist(x=x,y=f.x,type=type,main=main),args))
}
slider(refresh,
c("x limit (min)","x limit (max)",
"parameter 1","parameter 2")[1:length(mins)],
mins,maxs,deltas,defaults)
refresh()
cat("use sliders to select parameters!n")
}
Here are a few examples to show that the idea works:
• gamma distribution
26 show the density of the gamma distribution 26 ≡ ⊂ 30
slider.show.density("gamma",mins=c(0,0,.01,.01),
maxs=c(100,100,50,50),
deltas=c(.1,.1,.01,.01),
defaults=c(0,10,4,1), bty="n")
• beta distribution
27 show the density of the beta distribution 27 ≡ ⊂ 30
slider.show.density("beta",mins=c(0,0,0,0),
maxs=c(1,1,10,10),
deltas=c(.1,.1,.01,.01),
defaults=c(0,1,1,1),col="blue")
• binomial distribution
28 show the density of the binomial distribution 28 ≡ ⊂ 30
slider.show.density("binom",mins=c(0,0,1,0),
maxs=c(100,100,100,1),
deltas=c(1,1,1,.01),
defaults=c(0,20,5,0.5),type="h")
• Poisson distribution
29 show the density of the Poisson distribution 29 ≡ ⊂ 30
slider.show.density("pois",mins=c(0,0,.01),
maxs=c(100,100,50),
deltas=c(1,1,.1),
defaults=c(0,20,5),type="h")
The diﬀerent presenter functions could be collected in central function.
23

30 * 20 + ≡
slider.present.density<-function()
{
define slider.show.density 25
f0<- define slider.show.normal.density 23
f1<- function(...){ show the density of the gamma distribution 26 }
f2<- function(...){ show the density of the beta distribution 27 }
f3<- function(...){ show the density of the binomial distribution 28 }
f4<- function(...){ show the density of the Poisson distribution 29 }
slider(but.functions=c(f0,f1,f2,f3,f4),
but.names=c("show NORMAL","show GAMMA","show BETA",
"show BINOMIAL","show POISSON"))
}
slider.present.density()
Now the reader is prepared to take a further look at the implementations of the
slider functions of this paper.
4 Histograms and Density Traces
4.1 Class number of histograms – slider.hist
In this section we propose a function which allows you to study the effect of the
number of classes on the shape of a histogram. For this purpose a widget with a
slider controlling the number of classes is opened and a histogram is computed.
By moving the slider the user changes the number of classes and the histogram
is redrawn. The implementation uses two special features: 1.) Sometimes we
want to add further graphical elements like a density curve to the histogram.
For this job slider.hist has a slot to deliver a panel function. This panel
function will be called after updating the histogram. 2.) Additional graphics
parameter can be transfered by the "..." argument of slider.hist to the call
of hist which constructs the histogram.
Remarks: The default number of breaks is computed by hist. If a breaks
argument is found in the call of slider.hist it will be removed.
31 define slider.hist 31 ≡ ⊂ 2, 32, 33, 60
slider.hist<-function(x,panel=rug,...)
{
check if x is a vector 14
args<-list(...)
args<-args[names(args)!="breaks"]
ClassNumber<-length(hist(x,plot=FALSE)$breaks)
if(!any("main"==names(args)))args<-c(args,list(main=x.name))
xrange<-range(x); num<-slider(no=1)
breaks<-seq(xrange[1],xrange[2],length=num+1)
do.call("hist",c(alist(x=x,breaks=breaks),args))
panel(x)
}
slider(refresh,"ClassNumber",1,100,1,ClassNumber); refresh()
"use slider to select number of classes"
}
24

4.1.1 Test
Here is a check to test whether we can change the color.
32 check 32 ≡
slider.hist(log(islands),col="green")
To test the panel feature we add a normal density curve to the histogram.
33 check 32 + ≡
slider.hist(rivers,xlab="RIVERS",col="red",probability=TRUE,
pa=function(x){
}
)
4.2 Width and kernel of a density trace – slider.density
The main parameters of density traces are width and kernel type. In the function
slider.density() both of them are selected by sliders. The alternative idea to
implement the selection of the kernel function by buttons would have required
seven buttons. Therefore, a slider solution seems to be a little bit smarter. To
be able to add further graphical elements a panel function has been included.
Technical Remarks: 1.) The kernel type is stored in the slider variable kno. 2.)
In slider.density() we have two different effect functions where the second
one set.kernel() calls the first slider function (refresh()).
34 define slider.density 34 ≡ ⊂ 3, 35, 36, 60
slider.density<-function(x,panel=rug,...)
{
args<-list(...)
if(!any("main"==names(args))) args<-c(args,list(main=x.name))
kernel<-c("gaussian", "epanechnikov","rectangular",
"triangular","biweight", "cosine", "optcosine")
slider(obj.name="kno",obj.value=1)
width<-slider(no=1)*diff(range(x))/100
kno<-slider(obj.name="kno"); kernel<-kernel[kno]
xy<-density(x,width=width,kernel=kernel)
do.call("plot",c(alist(x=xy),args))
title(paste("nnwidth =",signif(width,4),", kernel =",kernel))
panel(x)
}
set.kernel<-function(...){
kernel<-slider(no=2)
slider(obj.name="kno",obj.value=kernel)
refresh()
25

}
bw.default<-diff(range(x))/density(x)$bw
slider(c(refresh,set.kernel),
c("width (% of range)","kernel"),
c(.1,1),c(100,7),c(.1,1),c(bw.default,1)
)
refresh()
cat("use slider to select width of window and to select kernel:n")
print(cbind("no"=1:7,"kernel"=kernel))
}
4.2.1 Test
At first we test slider.density without a panel function.
35 check of slider.density 35 ≡
slider.density(rivers,xlab="RIVERS",col="red")
Now we test a call of slider.density with a panel function to plot a fitted
normal density curve.
36 check of slider.density 35 + ≡
panel=function(x){
}
)
4.2.2 Help Page
37 define help of slider.hist 37 ≡
name{slider.hist}
title{interactive histogram and density traces}
alias{slider.hist}
alias{slider.density}
description{
The functions code{slider.hist} and code{slider.density}
compute histograms and density traces
whereas some parameter are controlled by sliders.
code{slider.hist} computes a histogram; the number of classes is
defined by a slider.
code{slider.density} computes a density trace; width and
type of the kernel are defined by sliders.
}
usage{
slider.hist(x, panel, ...)
26

slider.density(x, panel, ...)
}
arguments{
item{x}{ data set to be used for plotting }
item{panel}{ function constructing additional graphical elements to the plot}
item{dots}{ additional (graphics) parameters which are passed to
the invoked high level plotting function }
}
details{
code{slider.hist} draws a histogram of the data set code{x} by
calling code{hist} and opens a Tcl/Tk widget with one slider.
The slider defines the number of classes of the histogram. Changing the
slider results in redrawing of the plot. For further
details see the help page of code{hist}. code{rug} is used as the
default panel function.
code{slider.density} draws a density trace of the data set code{x}
by code{plot(density(...))} and opens a Tcl/Tk-widget with two
sliders. The first slider defines the width of the density trace
and the second one the kernel function:
code{"1-gaussian", "2-epanechnikov", "3-rectangular",
"4-triangular","5-biweight", "6-cosine", "7-optcosine"}
Changing one of the sliders results in a redrawing of the plot.
For further details see the help page of code{density}.
code{rug} is used as the default panel function.
}
value{
a message about the usage
}
references{ ~~ }
author{ Hans Peter Wolf }
seealso{ code{link{hist}}, code{slider}}
examples{
dontrun{
## This example cannot be run by examples() but should be work in an inter-
active R session
slider.hist(log(islands))
}
dontrun{
active R session
slider.density(rivers,xlab="rivers",col="red")
}
dontrun{
active R session
panel=function(x){
lines(xx,yy)
rug(x)
print(summary(yy))
}
)
}
27

}
keyword{ univar }
keyword{ iplot }
5 Brushing Functions
5.1 A draftsman’s display with Brushing – slider.brush.pairs
A draftsman’s display is a nice graphics to show two dimensional dependencies
of the variables of a multivariate data set. Brushing allows you to mark a subset
of the data points. The subset of the points is defined by the condition that the
coordinate of a selected variable has to lie within a fixed interval. By the function
slider.brush.pairs() the user can select a variable (dimension) as well as an
interval for recoloring the points satisfying a condition with color "red". The
function pairs() is not used in the refresh function avoiding computational
overhead. So the complexity of code is mostly caused by constructing the plot.
38 define slider.brush.pairs 38 ≡ ⊂ 4, 39, 60
slider.brush.pairs<-function(x,...)
{
args<-list(...)
# preparation of data
m<-dim(x)[2]; for(j in 1:m) x[,j]<-as.numeric(x[,j])
mins<-apply(x,2,min); maxs<-apply(x,2,max)
delta<-(maxs-mins)/100
# initial plot
varnames<-paste("var no",1:m,":",colnames(x))
dev.new(); par(mfrow=c(m,m),oma=c(0,0,0,0),mai=c(0,0,0,0),...)
usr.array<-array(0,c(m,m,4)); axes<-FALSE
for(i in 1:m){
for(j in 1:m){
# plot(x[,j],x[,i],axes=axes,type="p")
do.call("plot",c(alist(x=x[,j],y=x[,i],type="p"),args))
usr.array[i,j,] <- usr<-par()$usr
if(i==j) text(usr[1],usr[4],varnames[i],adj=c(0,1))
rect(usr[1],usr[3],usr[2],usr[4])
}
}
# update function
vmin<-slider(no=1)/100; vmax<-vmin+slider(no=2)/100
vno <-slider(no=3)
vmin<-mins[vno]*(1-vmin)+maxs[vno]*(vmin)
vmax<-mins[vno]*(1-vmax)+maxs[vno]*(vmax)
ind <-vmin<=x[,vno] & x[,vno]<=vmax
for(i in 1:m){
for(j in 1:m){
par(mfg=c(i,j),usr=usr.array[i,j,])
points(x[ ,j],x[ ,i],col=0,cex=2,pch=19)
points(x[ ind,j],x[ ind,i],col="red",pch=1)
points(x[!ind,j],x[!ind,i],col="blue",pch=19)
}
28

}
}
# slider definition
slider(refresh,
c("lower limit (% of range)","width (% of range)",
paste("variable no: 1 ..",m)),
c(0,0,1), c(100,100, m), c(1,1,1), c(0,30,1)
)
cat("use sliders to select variable and interval widthn")
}
5.1.1 Test
We will use the famous iris data to test the brushing function.
39 test of slider.brush.pairs 39 ≡
define slider.brush.pairs 38
usr.array<-slider.brush.pairs(iris,cex=.2)
5.2 Scatter plot brushing – slider.brush.plot.xy
The function slider.brush.plot.xy() computes an xy-plot and recolors a
data point "red" if the value of its third variable is in the fixed interval.
40 define slider.brush.plot.xy 40 ≡ ⊂ 5, 41, 60
slider.brush.plot.xy<-function(x,y=NULL,z=NULL,...)
{
check if x, y,z are vectors or x is matrix with >2 columns 16
args<-list(...)
if(!any("main"==names(args)))
args<-c(args,list(main=paste(x.name,"<-->",y.name)))
if(!any("xlab"==names(args)))args<-c(args,list(xlab=x.name))
if(!any("ylab"==names(args)))args<-c(args,list(ylab=y.name))
do.call("plot.default",c(alist(x=x,y=y,pch=19),args))
zrange<-range(z); z1<-slider(no=1); z2<-slider(no=2)
zmin<-z1; zmax<-z1+z2; ind<-zmin<=z&z<=zmax; pos<-par()$usr
rect(pos[2],pos[4],pos[1]*.5+pos[2]*.5,pos[3]*.1+pos[4]*.9,
col="white",border=NA)
txt<-paste(z.name,"(red) in [",format(zmin,digits=4),",",
format(zmax,digits=4),"]",sep="")
text(pos[2],pos[4],txt,adj=c(1,1),col="red",cex=0.7)
col<-c("black","red")[1+ind]
points(x,y,col=col,pch=19,
cex=if("cex" %in% names(args)) args$cex else 1)
}
z.min<-min(z); z.max<-max(z); delta<-(z.max-z.min)/100
reset<-function(...){
do.call("plot",c(alist(x=x,y=y,col="red",pch=19),args)); pos<-par()$usr #090216
rect(pos[2],pos[4],pos[1]*.4+pos[2]*.6,pos[3]*.1+pos[4]*.9,
col="white",border=NA)
}
29

slider(refresh,
c("minimum of z","interval width of z"),
c(z.min,0),c(z.max+delta,(z.max-z.min)+delta),
c(delta,delta),c(z.min-delta,(z.max-z.min)/2),
reset.function=reset
)
refresh()
cat("use sliders to select interval for inking pointsn")
}
5.2.1 Test
41 test of slider.brush.plot.xy 41 ≡
deﬁne slider.brush.plot.xy 40
# data<-matrix(rnorm(900), 300,3)
# slider.brush.plot.xy(data[,1],data[,2],sqrt(data[,1]^2+data[,2]^2),
# main="hallo")
# slider.brush.plot.xy(iris[,1],iris[,2],iris[,3])
slider.brush.plot.xy(iris[,1:3],cex=1.5)
5.2.2 Help Page
42 deﬁne help of slider.brush.pairs and of slider.brush.plot.xy 42 ≡
name{slider.brush}
title{interactive brushing functions}
alias{slider.brush.pairs}
alias{slider.brush.plot.xy}
description{
These functions compute a pairs plot or a simple xy-plot and
open a slider control widget for brushing.
code{slider.brush.pairs} computes a pairs plot; the user defines an
interval for one of the variables and in effect all data points
in this interval will be recolored.
code{slider.brush.plot.xy} computes an xy-plot; the user defines a
interval for a third variable code{z} and all points
code{(x,y)} will be recolored red if the code{z} value is in the inter-
val.
}
usage{
slider.brush.pairs(x, ...)
slider.brush.plot.xy(x, y, z, ...)
}
arguments{
item{dots}{ new settings for global graphics parameters }
item{x}{ matrix or data frame or vector }
item{y}{ vector of y values if code{x} is not a matrix }
item{z}{ vector of z values if code{x} is not a matrix }
}
30

details{
code{slider.brush.pairs} draws a pairs plot of the data set code{x}.
The first slider defines the lower limit of the interval and the
second its width. By the third slider a variable is selected.
All data points for which the selected variable is in the interval
are recolored red.
code{slider.brush.plot.xy} draws an xy-plot of the data set code{x}.
The first slider defines the lower limit of the interval of z values
and the second one its width. All data points for which the variable z
is in the interval are recolored red.
}
value{
}
references{ W. S. Cleveland, R. A. Becker, and G. Weil. The Use of
Brushing and Rotation for Data Analysis. InW. S. Cleveland
and M. E. McGill, editors, Dynamic Graphics for
Statistics. Wadsworth and Brooks/Cole, Pacific Grove,
CA, 1988. }
seealso{ code{link{pairs}}, code{link{plot}} }
examples{
dontrun{
active R session
slider.brush.pairs(iris)
}
dontrun{
active R session
slider.brush.plot.xy(iris[,1:3])
}
}
keyword{ iplot }
6 Time Series Plots
6.1 Splitted time series – slider.split.plot.ts
Often there is a periodical structure in time series of a fixed period. Then
you will like to have a tool to separate and compare the sections defined by
the season. The function slider.split.plot.ts lets you select the length of
a saison and one of the limits between two saisons. Then in the time series
plot fivenum summary statistics or a regression lines are added to each of the
sections.
43 define slider.split.plot.ts 43 ≡ ⊂ 6, 44, 60
slider.split.plot.ts<-function(x,type="l",...)
{
check if x is a time series 17
args<-list(...)
n<-length(x); xmin<-min(x); xmax<-max(x)
31

xdelta<-xmax-xmin
slider(obj.name="summary.type",obj.value="linear")
# initialization
summary.type<-slider(obj.name="summary.type")
width<-slider(no=1)
limit<-slider(no=2)
n.sec<-1
limit<-limit-width*ceiling((limit-xmin)/width)
# plot: # plot(x,y,type=type,bty="n",xlab="",ylab="")
do.call("plot",c(alist(x,y,type=type),args))
limit<-limit-width-width/n.sec; j<-0
# abline(v=limits,lwd=0.5,lty=3)
while(limit<xmax){ j<-j+1
limit<-limit+width/n.sec; limit2<-limit+width
ind<-limit<=x & x<=limit2
xx<-x[ind]; yy<-y[ind]; if(length(xx)<2) next
abline(v=limit,lwd=0.5,lty=3,col=1)
if(summary.type=="linear"){
coef<-lm(yy~xx)$coef
segments(limit, coef[1]+coef[2]*limit,
limit2,coef[1]+coef[2]*limit2,col=j)
}
if(summary.type=="five.number"){
five<-fivenum(yy)
xx<-0.5*(limit+limit2)
points(xx,five[3],pch=19,cex=1,col="red")
segments(xx,five[1],xx,five[2],lwd=3,col="red")
segments(xx,five[4],xx,five[5],lwd=3,col="red")
}
}
abline(v=slider(no=2),lwd=0.5,lty=1,col=1)
}
f1<-function(...){
slider(obj.name="summary.type",obj.value="linear")
refresh()
}
f2<-function(...){
slider(obj.name="summary.type",obj.value="five.number")
refresh()
}
slider(refresh,c("width of window","limit"),
c(xdelta/length(x)*3,xmin),c(xdelta,xmax),
c(xdelta/1000,xdelta/1000),c(xdelta/4,xmin),
but.functions=c(f1,f2),
but.names=c("linear model","fivenum summary")
)
refresh()
cat("select window and summary type and look at time series!n")
}
6.1.1 Test
The test call uses the time series co2.
32

44 test of slider.split.plot.ts 44 ≡
deﬁne slider.split.plot.ts 43
slider.split.plot.ts(co2,col="red",main="co2")
6.1.2 Help Page
45 deﬁne help of slider.split.plot.ts 45 ≡
name{slider.split.plot.ts}
title{interactive splitting of time series}
alias{slider.split.plot.ts}
description{
code{slider.split.plot.ts} plots linear fitted lines or
summary statistics in sections of a time series.
The sections are controlled by sliders.
}
usage{
slider.split.plot.ts(x, type="l", ...)
}
arguments{
item{x}{ time series or vector}
item{type}{ plotting type: code{type} will be forwarded to function code{plot}}
item{dots}{ additional graphics parameters }
}
details{
code{slider.split.plot.ts} draws a time series plot and let you define
sections of the series by fixing a limit on the time scale as well as
a window width.
The whole range of the series is partitioned in pieces of the same
length in a way that the fixed limit will be one of the section limits.
Then linear models are fitted and plotted in the sections.
Alternatively -- by pressing the button code{fivenum summary} --
summary statistics are drawn instead of the model lines.
The first slider fixes the width of the sections and
the second one the limit between two of them.
By clicking on button code{linear model} or code{fivenum summary}
the user switches between drawing model curves and five number summary.
}
value{
}
seealso{ code{link{plot}} }
examples{
dontrun{
active R session
slider.split.plot.ts(as.vector(sunspots)[1:100])
}
}
keyword{ iplot }
33

6.2 Zooming in Time Series – slider.zoom.plot.ts
Some times time series happen to be very long — for example sunspots has
a length of 2820. For exploring a section of a series it is helpful to have a
zooming function. The function slider.zoom.plot.ts allows you to define a
time window interactively and it computes a time series plot of the selected
region. As a second feature the function extracts the data of two windows and
computes two time series plots one below the other. With this feature you are
able to compare different regions of the series by eye.
46 define slider.zoom.plot.ts 46 ≡ ⊂ 7, 47, 48, 60
slider.zoom.plot.ts<-function(x,n.windows=1,...)
{
args<-list(...)
tmin<-1; tmax<-length(x)
# initialization
width <- slider(no=1); tstart1 <- slider(no=2); tend1 <- width+tstart1
if(n.windows>1){
tstart2 <- slider(no=3); tend2 <- width+tstart2
}
# plot
par(mfrow=c(2+(n.windows>1),1),mai=c(.5,0.5,.1,0))
# plot(x,y,type="l",bty="n",xlab="",ylab="")
do.call("plot",c(alist(x,y,type="l",bty="n"),args))
abline(v=c(x[tstart1],x[tend1]),col="red")
lines(x[tstart1:tend1],y[tstart1:tend1],col="red",lty=2)
if(n.windows>1){
abline(v=c(x[tstart2],x[tend2]),col="blue")
lines(x[tstart2:tend2],y[tstart2:tend2],col="blue",lty=3)
}
usr<-par()$usr
ind<-tstart1:tend1
plot(x[ind],y[ind],type="b",col="red",bty="n", # ylim=usr[3:4],
xlim=c(x[tstart1],x[tstart1]+width*diff(x[1:2])))
if(n.windows>1){
ind<-tstart2:tend2
plot(x[ind],y[ind],type="b",col="blue",bty="n", # ylim=usr[3:4],
xlim=c(x[tstart2],x[tstart2]+width*diff(x[1:2])))
}
par(mfrow=c(1,1))
}
if(n.windows<2){
slider(refresh,c("width of window","start of window"),
c(1,1),c(tmax,tmax),c(1,1),c(ceiling(tmax/4),1))
}else{
slider(refresh,
c("width of window","start window 1","start window 2"),
c(1,1,1),c(tmax,tmax,tmax),c(1,1,1),c(ceiling(tmax/4),1,ceiling(tmax/2)))
}
refresh()
cat("select window and look at time series!n")
}
34

6.2.1 Test
The ﬁrst test will be done with co2.
47 test of slider.zoom.plot.ts 47 ≡
slider.zoom.plot.ts(co2,2,main="co2")
For comparing two sections of a series take a look at sunspots.
48 test of slider.zoom.plot.ts 47 + ≡
slider.zoom.plot.ts(sunspots,2)
6.2.2 Help Page
49 deﬁne help of slider.zoom.plot.ts 49 ≡
name{slider.zoom.plot.ts}
title{interactive zooming of time series}
alias{slider.zoom.plot.ts}
description{
This function shows one or two sections of a time series. The window(s) is
(are) controlled by sliders.
}
usage{
slider.zoom.plot.ts(x, n.windows, ...)
}
arguments{
item{x}{ time series }
item{n.windows}{ code{if(n.windows>1} two sections are defined }
item{dots}{ additional graphical parameters }
}
details{
code{slider.zoom.plot.ts} plots the original time series and it lets you
select one or two sections of the series by fixing the width(s) and the
starting point(s) of the region(s). Then the section(s) of the series is (are)
plotted separately one below the other.
The first slider defines the width of the section(s).
The second (third) one sets the start of the first (second) section.
}
value{
}
examples{
dontrun{
active R session
slider.zoom.plot.ts(co2,2)
35

}
}
keyword{ iplot }
6.3 Smoothing Time Series – slider.smooth.plot.ts
Tukey has proposed a set of smoothing operations for time series. Some of them
are implemented by the function smooth(). slider.smooth.plot.ts supports
the user to find the suitable combination of smoothing operations. Interactively
he is able to smooth a time series step by step. The filter that should be used
in the next step is selected by Tcl/Tk control panel.
50 define slider.smooth.plot.ts 50 ≡ ⊂ 8, 51, 60
slider.smooth.plot.ts<-function(x,...)
{
t.x<-x; x<-y
args<-list(...)
kind<-c("3RS3R", "3RSS", "3RSR", "3R", "3", "S")
slider(obj.name="nts",obj.value=x)
slider(obj.name="kind",obj.value="3")
slider(obj.name="history.kind",obj.value="")
# initialization
choice<-slider(obj.name="kind"); print(choice)
history.kind<-slider(obj.name="history.kind")
history.kind<-c(history.kind,choice)
slider(obj.name="history.kind",obj.value=history.kind)
xx<-slider(obj.name="nts")
xx<-smooth(xx,kind=choice)
# plot(t.x,x,type="l",lty=3,bty="n",xlab="",ylab="")
do.call("plot",c(alist(t.x,x,type="l",lty=3,bty="n"),args))
title(paste(history.kind,collapse="->"))
points(t.x,xx,type="l",col="red") # lty=1+which(choice==kind))
slider(obj.name="nts",obj.value=xx)
}
reset<-function(...){
slider(obj.name="nts",obj.value=x)
slider(obj.name="kind",obj.value="3")
slider(obj.name="history.kind",obj.value="")
refresh()
}
f1<-function(...){slider(obj.name="kind",obj.value="3RS3R");refresh()}
f2<-function(...){slider(obj.name="kind",obj.value="3RSS"); refresh()}
f3<-function(...){slider(obj.name="kind",obj.value="3RSR"); refresh()}
f4<-function(...){slider(obj.name="kind",obj.value="3R"); refresh()}
f5<-function(...){slider(obj.name="kind",obj.value="3"); refresh()}
f6<-function(...){slider(obj.name="kind",obj.value="S"); refresh()}
refresh()
slider(but.functions=c(f1,f2,f3,f4,f5,f6),
but.names=kind,reset.function=reset)
cat("select type of smoothing a la Tukey and look the result!n")
}
36

6.3.1 Test
For testing we use some random numbers.
51 test of slider.smooth.plot.ts 51 ≡
deﬁne slider.smooth.plot.ts 50
slider.smooth.plot.ts(rnorm(100),main="nrandom",col="blue")
6.3.2 Help Page
52 deﬁne help of slider.smooth.plot.ts 52 ≡
name{slider.smooth.plot.ts}
title{interactive Tukey smoothing}
alias{slider.smooth.plot.ts}
description{
code{slider.smooth.plot.ts} computes smooth curves
of a time series plot by Tukey's smoothers.
The kind of smoothing is controlled by a Tcl/Tk widget.
}
usage{
slider.smooth.plot.ts(x, ...)
}
arguments{
item{x}{ time series }
item{dots}{ additional graphical parameters }
}
details{
code{slider.smooth.plot.ts} draws the time series code{x}.
The user selects a filter of the set
code{c("3RS3R", "3RSS", "3RSR", "3R", "3", "S")}
step by step and the resulting curve is added to the plot.
The selection is performed by pressing a button of the control
widget of code{slider.smooth.plot.ts}.
The button code{reset} restarts the smoothing process.
}
value{
}
references{ Tukey, J. W. (1977). Exploratory Data Analysis, Reading
Massachusetts: Addison-Wesley.
}
seealso{ code{link{plot}}, code{link{smooth}} }
examples{
dontrun{
active R session
slider.smooth.plot.ts(rnorm(100))
}
}
keyword{ iplot }
37

7 Regression
7.1 Smoothing by lowess – slider.lowess.plot
The result of the lowess smoother heavily depends on the setting of the param-
eter f. This parameter controls the smoother span and has to be fixed by the
user. The function slider.lowess.plot allows the user to select the span and
to observe the corresponding line. Furthermore, the number of iterations are
set interactively.
53 define slider.lowess.plot 53 ≡ ⊂ 9, 54, 60
slider.lowess.plot<-function(x,y=NULL,...)
{
# slider function to draw lowess smoother, pwolf 080525
check if x is a matrix or x and y are vectors 15
args<-list(...)
f<-slider(no=1)
iter<-slider(no=2)
xy<-lowess(x,y,f=f,iter=iter)
# plot(x,y,bty="n")
do.call("plot",c(alist(x,y,bty="n"),args))
lines(xy)
title(paste("nnlowess: f =",signif(f,4),", iterations =",iter))
lines(xy)
}
slider(refresh,
c("smoother span","iterations"),
c(.01,1),c(1,7),c(.01,1),c(2/3,3)
)
refresh()
cat("use slider to select smoother span!n")
}
7.1.1 Test
54 check of slider.lowess.plot 54 ≡
define slider.lowess.plot 53
slider.lowess.plot(cars,col="red")
7.1.2 Help Page
55 define help of slider.lowess.plot 55 ≡
name{slider.lowess.plot}
title{interactive lowess smoothing}
alias{slider.lowess.plot}
description{
38

code{slider.lowess.plot} computes an xy-plot of the data and
adds LOWESS lines. The smoother
span and the number of iterations are selected by sliders.
}
usage{
slider.lowess.plot(x, y, ...)
}
arguments{
item{x}{ data set to be used for plotting or vector of x values }
item{y}{ vector of y values in case code{x} is not a matrix }
item{dots}{ additional (graphics) parameter settings }
}
details{
code{slider.lowess.plot} computes a scatterplot of the data.
Then a LOWESS smoother line is added to the plot.
For more details about the lowess parameters code{f} and code{iter}
take a look at the help page of code{lowess}.
The parameters are set by moving sliders of the
control widget. The first slider defines the smoother span code{f}
and the second one the number of iterations.
}
value{
}
references{ for references see help file of code{lowess} }
seealso{ code{link{lowess}}, code{slider} }
examples{
dontrun{
active R session
slider.lowess.plot(cars)
}
}
keyword{ iplot }
7.2 Sensitivity of Regression Line – slider.bootstrap.lm.plot
To understand the senitivity of a regression curve the function slider.bootstrap.lm.plot
shows the user a lot of bootstrap regression lines computed on samples of the
data points. The number of repetitions, the random seed and the polynomial
degree is controlled by sliders.
56 deﬁne slider.bootstrap.lm.plot 56 ≡ ⊂ 10, 57, 60
slider.bootstrap.lm.plot<-function(x,y=NULL,...)
{
check if x is a matrix or x and y are vectors 15
args<-list(...)
n<-length(x)
ind<-order(x); x.orig<-x<-x[ind]; y.orig<-y<-y[ind]
# plot(x,y,...)
39

abline(lm(y~x),lwd=5)
# plot(x,y,...);
abline(coefyx<-lm(y~x)$coef, lwd=3)
polytype<-slider(no=1)
form<-paste(paste(sep="","I(x^",1:polytype,")"),collapse="+")
form<-as.formula(paste("y ~",form)); coef<-lm(form)$coef
yy<-outer(xx,0:polytype,"^")%*%coef; lines(xx,yy,lwd=2)
B<-slider(no=2); zz<-slider(no=3); set.seed(zz)
result<-matrix(0,1+polytype,B)
for(i in 1:B){
index<-sample(1:n,n,replace=TRUE)
x<-x.orig[index]; y<-y.orig[index]
coef<-lm(form)$coef
yy<-outer(xx,0:polytype,"^")%*%coef
lines(xx,yy,lwd=2,col=i,lty=2)
result[,i]<-coef
}
abline(coefyx, lwd=5)
result<-t(result);
colnames(result)<-c("intercept",paste(sep="","beta: x^",1:polytype))
print(summary(result))
}
slider(refresh,c("polynomial degree","number repetitions","random seed"),
c(1,1,1),c(5,50,100),c(1,1,1),c(1,10,1))
refresh()
"ok"
}
7.2.1 Test
57 test of slider..bootstrap.lm.plot 57 ≡
deﬁne slider.bootstrap.lm.plot 56
daten<-iris[,2:3]
7.2.2 Help Page
58 deﬁne help of slider.bootstrap.lm.plot 58 ≡
name{slider.bootstrap.lm.plot}
title{interactive bootstapping for lm}
alias{slider.bootstrap.lm.plot}
description{
code{slider.bootstrap.lm.plot} computes a scatterplot and
adds regression curves of samples of the data points.
The number of samples and the degree of the model are
controlled by sliders.
}
usage{
40

slider.bootstrap.lm.plot(x, y, ...)
}
arguments{
item{x}{ two column matrix or vector of x values if y is used }
item{y}{ y values if x is not a matrix }
item{dots}{ additional graphics parameters }
}
details{
code{slider.bootstrap.lm.plot} draws a scatterplot of the data points
and fits a linear model to the data set. Regression curves
of samples of the data are then added to the plot. Within a Tcl/Tk
control widget the degree of the model, the repetitions and the start
of the random seed are set. After modification of a parameter
the plot is updated.
}
value{
}
references{ ~~ }
examples{
dontrun{
active R session
daten<-iris[,2:3]
}
}
keyword{ iplot }
8 Appendix
8.1 Definition of slider
We will finish the paper by listing the definition of the function slider(). The
definition is copied from package relax.
59 define slider 59 ≡
slider<-function (sl.functions, sl.names, sl.mins, sl.maxs, sl.deltas,
sl.defaults, but.functions, but.names, no, set.no.value,
obj.name, obj.value, reset.function, title)
{
if (!exists("slider.env"))
slider.env <<- new.env(parent = .GlobalEnv)
if (!missing(no))
return(as.numeric(tclvalue(get(paste("slider", no, sep = ""),
env = slider.env))))
if (!missing(set.no.value)) {
try(eval(parse(text = paste("tclvalue(slider", set.no.value[1],
")<-", set.no.value[2], sep = "")), env = slider.env))
return(set.no.value[2])
}
if (!missing(obj.name)) {
if (!missing(obj.value))
assign(obj.name, obj.value, env = slider.env)
else obj.value <- get(obj.name, env = slider.env)
41

return(obj.value)
}
if (missing(title))
title <- "slider control widget"
if (missing(sl.names)) {
sl.defaults <- sl.names <- NULL
}
if (missing(sl.functions))
sl.functions <- function(...) {
}
require(tcltk)
nt <- tktoplevel()
tkwm.title(nt, title)
tkwm.geometry(nt, "+0+15")
assign("tktop.slider", nt, env = slider.env)
"relax"
for (i in seq(sl.names)) {
"relax"
eval(parse(text = paste("assign('slider", i, "',tclVar(sl.defaults[i]),env=slider.env)",
sep = "")))
tkpack(fr <- tkframe(nt))
lab <- tklabel(fr, text = sl.names[i], width = "25")
sc <- tkscale(fr, from = sl.mins[i], to = sl.maxs[i],
showvalue = TRUE, resolution = sl.deltas[i], orient = "horiz")
tkpack(lab, sc, side = "right")
assign("sc", sc, env = slider.env)
eval(parse(text = paste("tkconfigure(sc,variable=slider",
i, ")", sep = "")), env = slider.env)
sl.fun <- if (length(sl.functions) > 1)
sl.functions[[i]]
else sl.functions
if (!is.function(sl.fun))
sl.fun <- eval(parse(text = paste("function(...){",
sl.fun, "}")))
tkbind(sc, "<ButtonRelease>", sl.fun)
}
assign("slider.values.old", sl.defaults, env = slider.env)
tkpack(f.but <- tkframe(nt), fill = "x")
tkpack(tkbutton(f.but, text = "Exit", command = function() tkdestroy(nt)),
side = "right")
if (!missing(reset.function)) {
if (!is.function(reset.function))
reset.function <- eval(parse(text = paste("function(...){",
reset.function, "}")))
tkpack(tkbutton(f.but, text = "Reset", command = function() {
for (i in seq(sl.names)) eval(parse(text = paste("tclvalue(slider",
i, ")<-", sl.defaults[i], sep = "")), env = slider.env)
reset.function()
}), side = "right")
}
if (missing(but.names))
but.names <- NULL
for (i in seq(but.names)) {
but.fun <- if (length(but.functions) > 1)
but.functions[[i]]
else but.functions
if (!is.function(but.fun))
but.fun <- eval(parse(text = c("function(...){",
but.fun, "}")))
tkpack(tkbutton(f.but, text = but.names[i], command = but.fun),
side = "left")
}
invisible(nt)
}
42

Sliderfns

More Related Content

What's hot (20)

Viewers also liked (9)

Similar to Sliderfns (20)

More from rociogonzalezisabelpaz (16)

Recently uploaded (20)

Sliderfns