phi1 {BAS}

Compound Confluent hypergeometric function of two variables

Description

Compute the Confluent Hypergeometric function of two variables, also know as a Horn hypergeometric function or Humbert's hypergeometric used in Gordy (1998) with integral representation:

Usage

phi1(a, b, c, x, y, log = FALSE)

Arguments

a

a > 0

b

arbitrary

c

c > 0

x

x > 0

y

y > 0

log

logical indicating whether to return phi1 on the log scale

Details

phi_1(a,b,c,x,y) = [(Gamma(c)/Gamma(a) Gamma(a-c))] Int_0^1 t^(a-1) (1 - t)^(c-a-1) (1 - yt)^(-b) exp(x t) dt https://en.wikipedia.org/wiki/Humbert_series Note that Gordy's arguments for x and y are reversed in the reference above.

The original 'phi1' function in 'BAS' was based on 'C' code provided by Gordy. This function returns NA's when x is greater than 'log(.Machine$double.xmax)/2'. A more stable method for calculating the ‘phi1' function using R’s 'integrate' was suggested by Daniel Heemann and is now an option whenever $x$ is too large. For calculating Bayes factors that use the 'phi1' function we recommend using the 'log=TRUE' option to compute log Bayes factors.

Author(s)

Merlise Clyde ([e-mail])

Daniel Heemann ([e-mail])

References

Gordy 1998

See Also

Other special functions: hypergeometric1F1(), hypergeometric2F1(), trCCH()

Examples

# special cases
# phi1(a, b, c, x=0, y) is the same as 2F1(b, a; c, y)
phi1(1, 2, 1.5, 0, 1 / 100, log=FALSE)
hypergeometric2F1(2, 1, 1.5, 1 / 100, log = FALSE)

# phi1(a,0,c,x,y) is the same as 1F1(a,c,x)
phi1(1, 0, 1.5, 3, 1 / 100)
hypergeometric1F1(1, 1.5, 3, log = FALSE)

# use direct integration
phi1(1, 2, 1.5, 1000, 0, log=TRUE)