1
$\begingroup$

I am attempting to come up with a formula to describe something like the following:

Intended curve

I need the curve to start from 90% on the left side and it needs to hit 0% at the five year mark on the right. I'd like to be able to control how curved it is versus a straight line between the two points. What would be the best formula for something like this?

For the curious, this is intended to represent a tax proposal based on the length of time a stock is owned before it is sold, such that short term trades are highly penalized. I want to be able to tweak the curve for a couple different scenarios.

Thanks!

$\endgroup$
4
  • $\begingroup$ Just an idea, but what if you make the graph a quarter-circle (centered on the top-right)? $\endgroup$ Jun 11, 2015 at 16:02
  • $\begingroup$ @columbus8myhw And how would he control "how curved" it is? $\endgroup$
    – user228113
    Jun 11, 2015 at 16:06
  • $\begingroup$ A cubic spline might be a better idea. $\endgroup$ Jun 11, 2015 at 16:07
  • $\begingroup$ This is why we have differential equations.... $\endgroup$
    – Zach466920
    Jun 11, 2015 at 16:17

4 Answers 4

1
$\begingroup$

Here is a very simple solution:

$$f(x)=-\frac{90}{5^{2n+1}}(x-5)^{2n+1}$$

$n$ is the parameter of the curve. Below are the graphs for $\color{blue}{n=3}$ and $\color {red}{n=20}$:

enter image description here

$\endgroup$
0
$\begingroup$

Try $y=90 \left(1-\sqrt[n]{1-\left(1-\frac{x}{5}\right)^n}\right)$ for varying $n$ (which does not need to be an integer).

The curve is one quarter of a superellipse centered at $(5,90)$. It is tangent to the axes at the given points.

$\endgroup$
1
  • $\begingroup$ Any idea how I'd go about entering that into something like WolframAlpha's plotter? $\endgroup$ Jun 11, 2015 at 17:21
0
$\begingroup$

A nearly-correct solution would be to use an exponential decay $N = N_{0}e^{-t/\tau}$. To get this to hit zero at $t=5$, you can apply a correction by subtracting the value at $t=5$, $N_{0}e^{-5/\tau}$. A further correction to $N_{0}$ is then required to bring the start point back towards the required point.

These corrections are not quite perfect (the start point $N_{0}$ never quite returns to exactly where you want it)but lead to a curve that is ok for a show-curve.

For a curve from 90 at 0 to 0 at 5:

$$ 90(1+e^{-5/\tau})(e^{-x/\tau}-e^{-5/\tau}) $$

$\endgroup$
0
$\begingroup$

One possible curve:

$$y=\frac1{a(x+\sqrt{\frac{50}{9a}+\frac{25}4}-\frac52)}-\frac1{a(\sqrt{\frac{50}{9a}+\frac{25}4}+\frac52)},a>0$$

Methodology:

As equation $y=\frac1{ax},a>0$ has similar shape with the required curve, we may adopt it as an approximation measure. In the equation, larger the $a$, greater the "curvature". In order to make the $y=\frac1{ax}$ curve touching x-axis and y-axis, we shift the curve downward and leftward, and so we modify the equation into :

$$y+c=\frac1{a(x+d)}\text{ ,where }c,d>0$$

Sub x-intercept$=5$, y-intercept$=0.9$

$$0.9+c=\frac1{ad}$$ $$c=\frac1{a(d+5)}$$

Solving,

$$c=\frac1{a(\sqrt{\frac{50}{9a}+\frac{25}4}+\frac52)}$$$$d=a(\sqrt{\frac{50}{9a}+\frac{25}4}-\frac52)$$

$\endgroup$

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.