Null hypothesis μ_0.
When graphing a confidence interval, mean0 will be used for
xbar should xbar itself have the value NA.
mean1
Alternative hypothesis μ_1.
xbar
Observed ar{x}.
sd
Standard deviation in the data scale σ for normal-, or s for t-distribution.
df
Degrees of freedom for t-distribution.
n
Number of observations per group.
main, xlab, ylab, xlim, ylim, sub
Standard xyplot arguments. Default values
are constructed if these arguments are missing. The input value
main=NA
forces a new constructed main instead of using the main coming in through the
htest methods.
...
Additional xyplot arguments.
number.vars
Number of variables. 1 for a one-sample test, 2 for two-sample tests
and paired tests.
alpha.left, alpha.right
For type="hypothesis",
the sum of these two numbers is the probability of the Type I Error
α. When both of these numbers are positive, there is a two-sided
test. Note that it is not required that they be equal. If one of the
numbers is 0, then it is a one-sided test. For
type="confidence", 1 minus the sum of these two numbers is the
confidence level.
float
Logical. If TRUE, then the probabilities α, β, power, and
p-values or the confidence value are displayed on the graph. If FALSE, these values
are not displayed.
ntcolors
Vector of colors used in the graph. The default value is
"original", which implies the ten colors c(col.alpha = "blue", col.notalpha = "lightblue",
col.beta = "red", col.power = "pink", col.pvalue = "green",
col.pvaluetranslucent = green127, col.critical = "gray50",
col.border = black127, col.text = "black", col.conf = "lightgreen").
An alternative is "stoplight" which implies the ten colors
c(col.alpha = "red", col.notalpha = "honeydew2",
col.beta = "orange", col.power = "pink", col.pvalue = "blue",
col.pvaluetranslucent = blue127, col.critical = "gray50", col.border
= black127, col.text = "black", col.conf = "lightgreen").
The partially transparent colors are: black127="#0000007F",
green127="#00FF007F", blue127="#0000FF7F".
The user can enter any color scheme by specifying a vector of ten
named colors. The names are:
col.alpha, col.notalpha, col.beta, col.power, col.pvalue,
col.pvaluetranslucent, col.critical, col.border, col.text, col.conf.
digits.axis, digits.float, digits
digits.axis is the number of significant digits for the top
axis. digits.float is the number of significant digits for
the floating probability values on the graph. digits is a
convenience argument to set both digits.axis and
digits.float at the same time. These number is passed to the
format function.
distribution.name
Name of distribution.
type
"hypothesis" for a Hypothesis Test graph, or "confidence" for a
Confidence Interval graph.
zaxis, z1axis
Logical or list. Should the z-axis centered on μ_0, or the
z_1-axis centered on μ_1, be displayed? The list
version of the argument must have two components at and labels as
specified in panel.axis.
cex.z, cex.prob, cex.top.axis, cex.main
cex.z is the cex value for the z and z_1
axes on the plot. cex.prob is the cex value for the
floating probabilities on the graph. cex.top.axis is the cex value
for the top axis values. cex.main is the cex value for
the main title.
key.axis.padding
tuning constant to create additional room above the
graph for a larger cex.main to fit.
prob.labels
logical. If TRUE label the floating
probability values with their name, such as α. If
FALSE,
then don't label them. The default is TRUE for
type="hypothesis"
and FALSE for type="confidence".
xhalf.multiplier, yhalf.multiplier
Numerical tuning constants to control the width and height of the floating
probability values. Empirically, we need a smaller value for the
shiny
app then we need for direct writing onto a graphic device.
NTmethod
Character string used when shiny=TRUE. It is
normally calculated by the methods. NTmethod tells
shiny how to use or ignore the df and n
sliders.
"htest" objects by default are interpreted
as a single observation (n=1) of a t-statistic with
df degrees of freedom. The slider will let the user change
the df, but not the n.
"power.htest" objects are interpreted as a set of n
obervations per group and df is calculated as (n-1) for
single-sample tests and as 2(n-1) for two-sample tests.
The slider will let the user change n and will calculate the
revised df.
For the normal approximation to the binomial
(distribution.name="binomial"),
only n is meaningful. The df is always ignored.
For the default situation of t, determined by the initially
specified sample size n>1, the degrees of freedom is
calculated as (n-1) for single-sample tests and as
2(n-1) for two-sample tests. The default z, is
initially specified by a sample size n=1.
In all cases except the "binomial", the user can change the
interpretation of the n and df sliders. The
interpretation when both n and df are under user
control is not always obvious.
power, beta
Logical. If TRUE, then display that graph,
else don't display it. Passed forward to
powerplot.
Details
The graphs produced by this single function cover most of the first semester
introductory Statistics course. The htest method plots the
results of the stats::t.test function.
NormalAndTplot is built on xyplot.
Most of the arguments detailed in xyplot documentation work to
control the appearance of the plot.
Value
"trellis" object.
Note
This function is built on lattice and latticeExtra.
It supersedes the similar function
normal.and.t.dist built on base graphics that is used in many
displays in the book by Erich Neuwirth and me: R through Excel, Springer
(2009).
http://www.springer.com/978-1-4419-0051-7. Many details,
particularly the
alternate color scheme and the concept of floating probability labels,
grew out of discussions that Erich and I have had since the book was
published.
The method for "htest" objects incorporates ideas that Jay Kerns and I developed at the 2011 UseR! conference.
This version incorporates some ideas suggested by Moritz Heene.