Skip to contents

Directional information

Source: R/directional-information.R
directional_information.Rd

Quantifies the information about direction in a posterior distribution based on the directional probability. This function calculates such information using the difference in the probability of direction (see probability_direction()), after converting each probability to bits (also see svalue().

Usage

directional_information(
  x,
  side = "median",
  threshold = 0,
  threshold_split = "proportional",
  na_rm = FALSE
)

Arguments

x

A numeric vector of MCMC values.

side

A string indicating whether to calculate the directional information relative to the left side ("left"; x < threshold), or the right side ("right"; x > threshold). Positive information suggests greater evidence for the specified side. Defaults to "median", which uses the side of the median of x via direction().

threshold

A number of the threshold value.

threshold_split

A string indicating how to deal with threshold values:

  • "left" to include them on the left side,

  • "right" to include them on the right side,

  • "equal" to split them equally between the left and side,

  • "proportional" (default) to split them between the left and right sides proportionally to the values of x on the left and right sides,

  • "exclude" to drop the values of x equal to threshold (identical to using "proportional").

na_rm

A flag specifying whether to remove missing values.

Value

A number indicating the directional information in bits. If x has NA values but na_rm is FALSE, returns NA_real.

References

Kery, M., and Schaub, M. 2011. Bayesian population analysis using WinBUGS: a hierarchical perspective. Academic Press, Boston. Available from https://www.vogelwarte.ch/en/research/population-biology/book-bpa/.

See also

Other summary: direction(), kurtosis(), lower(), probability_direction(), pvalue(), pzeros(), skewness(), svalue(), upper(), variance(), xtr_mean(), xtr_median(), xtr_sd(), zeros(), zscore()

Examples


directional_information(0)
#> [1] 0
directional_information(1) # one coin flip of information
#> [1] 1
directional_information(c(1, 1)) # two coin flips
#> [1] 2
directional_information(c(1, 1, -1)) # x[2] and x[3] cancel out
#> [1] 1
directional_information(c(1, 1, -1, -1)) # both sides cancel out
#> [1] 0
directional_information(rnorm(1e3, mean = 0))
#> [1] 0.0577155
directional_information(rnorm(1e3, mean = 1))
#> [1] 2.328864
directional_information(rnorm(1e3, mean = 10)) # all coin flips are positive
#> [1] 1000
directional_information(rnorm(1e3, mean = -10)) # all coin flips are negative
#> [1] 1000
directional_information(rnorm(1e3, mean = 1e3)) # only quantiles matter
#> [1] 1000
directional_information(rnorm(1e6, mean = 1e3)) # more `x` implies more info
#> [1] 1e+06