0
$\begingroup$

I have following function $$f(x_0,x_1)=-x_1Ei(-a-\frac{bx_1}{x_0})\exp(a+\frac{bx_1}{x_0})$$. Where $a,b$ are greater than zero, $x_0,x_1$ can have positive values and $Ei(x)$ is the exponential integral. Now its perspective function is $$f_1(x_0)=-Ei(-a-\frac{b}{x_0})\exp(a+\frac{b}{x_0})$$ If I do following transformation $\frac{1}{t}=a+\frac{b}{x_0}$ then I have $$f_2(t)=-Ei(-\frac{1}{t})\exp(\frac{1}{t})$$ Now $f_2(t)$ is concave with respect to $t$. Based on this I assume that $f_1(x_0)$ is concave with respect to $x_0$. Since $f_1(x_0)$ is the perspective function of $f(x_0,x_1)$ therefore I conclude that $f(x_0,x_1)$ is jointly concave with respect to $(x_0,x_1)$. But when I fix $x_0$ and try to prove the concavity of $f(x_0,x_1)$ with respect to $x_1$ through double derivative test. Then I am unable to prove this. Is the function $f(x_0,x_1)$ not jointly concave because if it were then it should be concave with respect to $x_1$ also. Any help in this regard will be much appreciated. Thanks in advance.

$\endgroup$
1
$\begingroup$

\begin{align}f_1(x_0)&=-Ei(-a-\frac{b}{x_0})\exp(a+\frac{b}{x_0})\\ &=-Ei\left(\color{red}- \frac1t\right) \exp\left(\frac1t\right)\end{align}

$\endgroup$
6
  • $\begingroup$ I am extremely sorry. It was my typing mistake (I have corrected it, Do you agree that the transformation used in my post is right). It is actually, as you mentioned, a negative sign. So $-Ei(-\frac{1}{t})\exp(\frac{1}{t})$ is a concave function. Since $f_1(x_0)=f_2(t)$ so $f_1(x_0)$ is also concave and therefore $f(x_0,x_1)$ should be concave with respect to $x_1$ also. But I am unable to prove it through double derivative test. $\endgroup$ – Frank Moses Feb 5 '18 at 1:50
  • $\begingroup$ I am curious about my transformation $\frac{1}{t}=a+\frac{b}{x_0}$. Do you think it is right transformation? $\endgroup$ – Frank Moses Feb 5 '18 at 1:51
  • 1
    $\begingroup$ I haven't verified, but do you have a proof that if $g(t)$ is concave in $t$, then by a transformation of $\frac1t=a+\frac{b}{x_0}$, the function is concave in $x_0$? $\endgroup$ – Siong Thye Goh Feb 5 '18 at 1:58
  • $\begingroup$ No I do not have a proof. That's why I asked this question. But I describe my understanding below. With $\frac{1}{t}=a+\frac{b}{x_0}$ we have $t=\frac{x_0}{ax_0+b}$ therefore for the range of $x_0$ (from $0\to \infty$) the range of $t$ is $0 \to \frac{1}{a}$ and each individual value of $x_0$ produces a unique value of $t$. Since the function $f_2(t)$ is concave overall values of $t$ (which I can prove) therefore I think it is also concave over $0<t<\frac{1}{a}$ (Please correct me if I am wrong in this). Thanks in advance. $\endgroup$ – Frank Moses Feb 5 '18 at 2:09
  • 1
    $\begingroup$ Suppose $f_2(t)=-t$, it is a concave function, but $h(x) = f_2 \left( \frac{x}{x+1} \right) = -\frac{x}{x+1}$ is not a concave function in $x$. $\endgroup$ – Siong Thye Goh Feb 5 '18 at 2:17

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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