Using logarithmic derivative functions for assessing the risky weighting function for binary gambles (Q2437214): Difference between revisions

The cummulative prospect theory representation is used as a form of utility for a binary gamble \( U(G) = w(p)u(V_1) + (1 - w(p))u(V_2)\), where \(V_1 > V_2\) are outcome values, \(p \in [0,1]\), \(w(p), 1 - w(p)\) are risky weighting coefficients depending on probability \(p\) and it is additionally assumed that \(w(0) = 0\), \(w(1) = 1\) and \(w(p)\) is a differentiable increasing function of \(p\). The logarithmic derivative of a function \(g(x)\) is defined as follows: \[ \eta(x) = d(\ln g(x))/dx = g'(x)/g(x). \] Properties of the logarithmic derivative are studied and the function \(w(p)\) is examined in terms of its logarithmic derivative \(\eta(p) = w'(p)/w(p)\). Theoretical foundations are established that link how data can be obtained to estimate \(\eta(p)\). An experiment is derived that estimates \(\eta(p)\) across a range of probability values. The paper is concluded with a general discussion of the most suitable models for the function \(w(p)\).