The mode is a measure of central tendency. It is the value that occurs most frequently, or in a continuous probability distribution, it is the value with the most density. A distribution may have no modes (such as with a constant, or in a uniform distribution when no value occurs more frequently than any other), or one or more modes.
is.amodal(x, min.size=0.1)
is.bimodal(x, min.size=0.1)
is.multimodal(x, min.size=0.1)
is.trimodal(x, min.size=0.1)
is.unimodal(x, min.size=0.1)
Mode(x)
Modes(x, min.size=0.1)This is a vector in which a mode (or modes) will be sought.
This is the minimum size that can be considered a mode, where size means the proportion of the distribution between areas of increasing kernel density estimates.
The is.amodal function is a logical test of whether or not
x has a mode. If x has a mode, then TRUE is
returned, otherwise FALSE.
The is.bimodal function is a logical test of whether or not
x has two modes. If x has two modes, then TRUE
is returned, otherwise FALSE.
The is.multimodal function is a logical test of whether or not
x has multiple modes. If x has multiple modes, then
TRUE is returned, otherwise FALSE.
The is.trimodal function is a logical test of whether or not
x has three modes. If x has three modes, then TRUE
is returned, otherwise FALSE.
The is.unimodal function is a logical test of whether or not
x has one mode. If x has one mode, then TRUE
is returned, otherwise FALSE.
The Mode function returns the most frequent value when x
is discrete. If x is a constant, then it is considered amodal,
and NA is returned. If multiple modes exist, this function
returns only the mode with the highest density, or if two or more
modes have the same density, then it returns the first mode found.
Otherwise, the Mode function returns the value of x
associated with the highest kernel density estimate, or the first
one found if multiple modes have the same density.
The Modes function is a simple, deterministic function that
differences the kernel density of x and reports a number of
modes equal to half the number of changes in direction, although the
min.size function can be used to reduce the number of modes
returned, and defaults to 0.1, eliminating modes that do not have at
least 10% of the distributional area. The Modes function
returns a list with three components: modes, modes.dens,
and size. The elements in each component are ordered according
to the decreasing density of the modes. The modes component is
a vector of the values of x associated with the modes. The
modes.dens component is a vector of the kernel density
estimates at the modes. The size component is a vector of the
proportion of area underneath each mode.
The IterativeQuadrature,
LaplaceApproximation, and VariationalBayes
functions characterize the marginal posterior distributions by
posterior modes (means) and variance. A related topic is MAP or
maximum a posteriori estimation.
Otherwise, the results of Bayesian inference tend to report the
posterior mean or median, along with probability intervals (see
p.interval and LPL.interval), rather than
posterior modes. In many types of models, such as mixture models, the
posterior may be multimodal. In such a case, the usual recommendation
is to choose the highest mode if feasible and possible. However, the
highest mode may be uncharacteristic of the majority of the posterior.
IterativeQuadrature,
LaplaceApproximation,
LaplacesDemon,
LPL.interval,
p.interval, and
VariationalBayes.
# NOT RUN {
library(LaplacesDemon)
### Below are distributions with different numbers of modes.
x <- c(1,1) #Amodal
x <- c(1,2,2,2,3) #Unimodal
x <- c(1,2) #Bimodal
x <- c(1,3,3,3,3,4,4,4,4,4) #min.size affects the answer
x <- c(1,1,3,3,3,3,4,4,4,4,4) #Trimodal
### And for each of the above, the functions below may be applied.
Mode(x)
Modes(x)
is.amodal(x)
is.bimodal(x)
is.multimodal(x)
is.trimodal(x)
is.unimodal(x)
# }
Run the code above in your browser using DataLab