0
$\begingroup$

I would like to manipulate a functional equation where $F$ is a cumulative distribution function, which I would let Mathematica know.

Using a simple example, consider

$$(\rho + \lambda\overline{F}(w))V(w) = w + \lambda \int_w^\overline{w} V(x) \,\text{d}F(x)$$

where the integral is an expectation for the draws larger than $w$ (which is distributed according to $F$ over a support $[\underline{w}, \overline{w}]$) and $\overline{F}(w) = 1-F(w)$. I would like to manipulate this expression using the properties of the CDF, and eventually solve it for a function such as $V(w)$ —- in this case a simple task, but I would like Mathematica to do it.

I think RSolve should be the right tool to handle functional equations, but I can't manage to obtain an output. I am trying something like:

RSolve[(rho + l*(1-F[w]))*V[w] == w + l*Integrate[V[x], {F[x], w, w1}], V[w], w]
$\endgroup$
2
  • $\begingroup$ RSolve is not the right tool for this. Mathematica doesn't really have that much support for integro/functional equations. $\endgroup$
    – flinty
    Commented Nov 20, 2020 at 13:37
  • 1
    $\begingroup$ Maybe take derivative wrt w and convert to an ODE? $\endgroup$ Commented Nov 20, 2020 at 14:53

1 Answer 1

2
$\begingroup$

Under the assumption that the PDF exists we have $\int_w^{w1} V(x)\,dF(x)= \int_w^{w1} V(x)F'(x)\,dx$. Differentiating both sides of the equation under consideration by w, we reduce the integral equation under consideration to a usual ODE and solve it.

D[(rho + l*(1 - F[w]))*V[w], w] == D[w + l*Integrate[V[x]*F'[x], {x, w, w1}], w]
DSolve[%, V[w], w]
(*{{V[w] ->  C[1] + Inactive[Integrate][1/(l + rho - l F[K[1]]), {K[1], 1, w}]}}*)

No properties of the CDF are used in the above.

$\endgroup$
2
  • $\begingroup$ Thank you. This is a very useful starting point. I was wondering if there is a way to assume the behavior of F on the support extremes: F[wmax]=1 and F[wmin]=0. I tried to play around with $Assumptions but I could not figure it out. $\endgroup$
    – FDP
    Commented Nov 20, 2020 at 18:44
  • $\begingroup$ @FDP: The finiteness of the support of the CDF implies that V[w] is a linear function if w is bigger than the upper bound of the support. This conclusion is math, not Mathematica. $\endgroup$
    – user64494
    Commented Nov 20, 2020 at 18:55

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.