Finding the Equation of a Curve
Date: 10/02/2000 at 08:23:14 From: Erik Subject: Equation of a curve Dr. Math, I'm hoping you can help me find the equation to a certain curve. I have looked through several Internet sites listing graphs of common curves, but can't find the one I'm looking for. The curve I would like to describe mathematically starts at some point on the positive y-axis, grows increasingly positive in the x-direction, and perhaps decreases in y only slightly. At some point in the x-direction, the slope of the curve becomes sharply negative and approaches the x-axis. The curve should flatten out as it approaches the x-axis, which is an asymptote for the curve. I'll try to provide an illustration: |* * * * * * * | * * * * * | * | * | * | * | * | * | * | * | * | * | * * * * * * * +-------------------------------------------------- Generally, I would like to know how one might go about finding the equation of any curve. What process does one go through to find an equation for a simple curve like this one? Thanks for the help, Erik
Date: 10/02/2000 at 14:30:08 From: Doctor Ian Subject: Re: Equation of a curve Hi Erik, There is no 'standard' way to figure out what kind of function corresponds to a given graph, except to think of the different kinds of functions that you know about and see whether any of them are approximately the right shape. When you can't come up with anything off the top of your head, there are three other ways to proceed: 1) Think about whether the graph describes the behavior of any physical object or process that you know about. For example, an inverse-square rule (like the one that governs gravity and electrostatic processes) falls off asymptotically to zero with increasing distance, which describes one part of the curve, but it rises very quickly as distance approaches zero, so it doesn't describe the other part. 2) Look for two or more functions, each of which matches one part of the curve, but goes to zero everywhere else. One thing to look out for is a curve that has the right shape, but is shifted slightly, or inverted in some way. In your case, you might be able to find a curve that looks like this: * * | * | *| -------------------+--------------- |* | * | * | * If you move this up and to the right, you get something that looks a lot like your curve. Or something like: * | * | * |* -------------------+--------------- *| * | * | * | Which looks like your curve inverted and shifted. Now the part on the right looks like a logarithmic function. Unfortunately, log(x) isn't defined for negative x. But this is the general idea. A lot depends on what you want to do with the mathematical description. If you're trying to describe a causal relationship, then you want to hold out for a single function. But if you just want to replace a lookup with a computation, then it might suffice to find a polynomial that has the right shape over the interval of interest, without worrying about what it will look like outside that interval. For example, you might be able to find a cubic equation that looks like this: * * * * * * * * * * * * * * * |---------------------------| In a limited domain, this mimics your curve, which might be good enough. Having said all that, the curve you've drawn looks a lot like the right half of a normal distribution curve, pictured here in the Internet Glossary of Statistical Terms: Normal Distribution http://www.animatedsoftware.com/statglos/sgnormdi.htm and even more like the mirror image of a cumulative distribution function for a normal distribution. See section 0.2 in The Thickets Approach to P-Bounds http://class.ee.iastate.edu/berleant/home/me/vita/papers/thicketsChapter.htm for a side-by-side look at graphs of these two functions. To turn cumulative distribution function, F(X), into your curve, you would plot y = 1 - F(x) instead of just y = F(x). If you'd like to write back and say a little more about where you got the curve you're trying to describe, and what you plan to do with the description, I could be of more help. - Doctor Ian, The Math Forum http://mathforum.org/dr.math/
Date: 10/04/2000 at 17:04:06 From: Erik Subject: Re: Equation of a curve Dr. Ian, Thanks a lot for the very detailed and thorough explanation! Now that you mention it, the mirror of the cumulative normal distribution does look like what I am trying to describe very well. As far as what I am trying to do, I would like to find a way to describe the way people store and discard pending information. For instance, if the sitting president happens to have impeachment charges brought against him, an interested party will have this fact in the forefront of his mind for the time the impeachment is announced until the time the hearing is held (the flat part of the curve.) After the impeachment vote is taken and the results announced, the psychological impact of the impeachment proceedings will fall off fairly quickly but will probably never reach 0 (the negative slope and the asymptote on the x-axis.) Memory researchers usually model the memory decay portion of the curve as x^2, where 0 < x < 1 and f(x) represents a "memory percentage" of sorts. This fits the decay to the asymptote part of the curve, but not the flat part of the curve. I have looked on the Internet for some model that would be similar to the one I have described, but have not been able to find anything similar. I did find one site about mathematical modeling of human memory, but these models seemed to be dealing much more with a human's ability to remember specific facts for a given amount of time (which may be a little different from what I would like to measure -- the emotional impact of a pending fact). Has this explanation been clear? The cumulative normal distribution's mirror does seem to fit the general shape, but I'm not sure how to control the length of the relatively flat portion near the y-axis (i.e., how to distinguish between an event that is to be pending for 3 units of time versus one that is to be pending for 4 units of time). I'm also not really clear as to what the mathematical formula would be for this function. "The mirror of the cumulative normal distribution" makes sense to me, but I don't know what this would look like in an equation. Thanks again for the help. I appreciate you spending brainpower on this funny little problem! I also appreciated your general comments about how to figure what an equation should look like and was happy that I did the 'right' thing when I first started this problem (I just wish I'd thought of the normal distribution myself...) Best regards, Erik
Date: 10/05/2000 at 17:57:52 From: Doctor Ian Subject: Re: Equation of a curve Hi Erik, I'm glad I was able to help. One of the URLs that I mentioned had an equation for the normal distribution function; the cumulative distribution function is the integral of that function. And the mirror would be one minus the integral. But you've hit on the crux of the problem, which is that the flat part of the curve can stretch out indefinitely. This is the case (I would argue) because the person is continually being reminded of the pending event. So what you're trying to model is something like the charge of a battery that can sit in a recharger indefinitely before being used. There are two different behaviors (being charged, and discharging), so it shouldn't be a big surprise to find that it's difficult to find one equation that describes them both. I had fun thinking about the problem, and for what it's worth, I was about to send the general part of the answer (i.e., how to go about solving a problem like this one), when the similarity to the distribution function finally occurred to me. Once you see it, you wonder how you could have missed it, but I wouldn't say it's obvious at all. (I've shown it to a few other people, and none of them recognized it right away, either. And if you look at the form of the normal distribution function, it becomes clear why you wouldn't be likely to arrive at it by playing around symbolically.) Anyway, thanks for writing. Write back if you have any other questions, about this or anything else. - Doctor Ian, The Math Forum http://mathforum.org/dr.math/
Search the Dr. Math Library:
Ask Dr. MathTM
© 1994- The Math Forum at NCTM. All rights reserved.