Skip to main content

Table 1 Different Parametrizations: \(\mathcal {P}(K)\leftrightarrows \mathcal {P}(T)\) with β=α, α=θ α·δ/(1−α), λ=(1−α)/δ. \(\mathcal {P}(H)\leftrightarrows \mathcal {P}(T)\leftrightarrows \mathcal {P}(P)\) using \(\delta =\mu \left (\frac {1-\alpha }{\mu \nu ^{2}}\right)^{1-\alpha }\), \(\gamma =\left [\frac {\mu \cos (\pi \alpha /2)}{\alpha }\right ]^{\frac {1}{\alpha }}\left [\frac {1-\alpha }{\mu \nu ^{2}}\right ]^{\frac {1-\alpha }{\alpha }}\) and \(\theta =\frac {1-\alpha }{\mu \nu ^{2}}\)

From: pTAS distributions with application to risk management

Reference

Parameter

Abbreviation

Hougaard (1986)

(α,δ,θ)

\(\mathcal {P}(H)\)

Tweedie (1984)a, Haas and Pigorsch (2009)

(α,γ,δ)

\(\mathcal {P}(T)\)

Palmer et al. (2008)

(α,μ,ν)

\(\mathcal {P}(P)\)

Küchler and Tappe (2011)

(β,α,λ)

\(\mathcal {P}(K)\)

  1. a Tweedie (1984) considered exponential dispersion models with variance function of the form \(V(\mu)\propto \mu ^{p}\) such models with p>2 coincide with the pTAS family with α=(p−2)/(p−1), see Jørgensen (1987), chapter 4, for a detailed account