logarithmicmean_byname: Name- and element-wise logarithmic mean of matrices

In MatthewHeun/byname: An Implementation of Matrix Mathematics that Respects Row and Column Names

Name- and element-wise logarithmic mean of matrices

Description

The logarithmic mean of corresponding entries of a and b is 0 if a = 0 or b = 0, a if a = b, or (b - a) / (log(b) - log(a)) otherwise.

Usage

logarithmicmean_byname(a, b, base = exp(1))

Arguments

a	first operand (a matrix or constant value or lists of same).
b	second operand (a matrix or constant value or lists of same).
base	the base of the logarithm used when computing the logarithmic mean. (Default is base = exp(1).)

Details

This function performs a union and sorting of row and column names prior to performing logarithmic mean. Zeroes are inserted for missing matrix elements.

Internally, the third condition is implemented as (b - a) / log(b/a).

Note that (b - a) / log(b/a) = (a - b) / log(a/b), so logarithmic mean is commutative; the order of arguments a and b does not change the result.

Value

A matrix representing the name-wise logarithmic mean of a and b.

Examples

library(dplyr)
m1 <- matrix(c(1:6), nrow = 3, ncol = 2) %>% 
  setrownames_byname(c("r1", "r2", "r3")) %>% setcolnames_byname(c("c1", "c2")) %>% 
  setrowtype("row") %>% setcoltype("col")
m2 <- matrix(c(7:12), nrow = 3, ncol = 2) %>% 
  setrownames_byname(c("r2", "r3", "r4")) %>% setcolnames_byname(c("c2", "c3")) %>% 
  setrowtype("row") %>% setcoltype("col")
logarithmicmean_byname(m1, m2)
# This also works with lists
logarithmicmean_byname(list(m1, m1), list(m2, m2))
DF <- data.frame(m1 = I(list()), m2 = I(list()))
DF[[1,"m1"]] <- m1
DF[[2,"m1"]] <- m1
DF[[1,"m2"]] <- m2
DF[[2,"m2"]] <- m2
logarithmicmean_byname(DF$m1, DF$m2)
DF %>% mutate(logmeans = logarithmicmean_byname(m1, m2))

MatthewHeun/byname documentation built on Feb. 17, 2024, 4:51 p.m.