Thread: Piecewise function problem.

As I understand it, you would be helped by two functions, call them g(x) and h(x), where g(x) starts at g(0) = 1 and tends toward 0 as x --> infinity; and h(x) starts at h(0) = 0 and tends toward 1 as x --> infinity. Then you would be able to build e^(f(x)*g(x)) + x*h(x). This would start off like y = e^(f(x)) and smoothly approximate y = x + 1. Not knowing any particular values you want for this graph, we can get a family of approximately desired functions by taking g(x) = e^(-x) and I'm still working on a good answer for h(x), but maybe you can come up with your own, or tell me whether I'm understanding what you want correctly.

Maybe h(x) = 1/(x+1).

If you want to delay/hasten the on-set of the effects of g(x) or h(x) you can either shift them by g(x + a), h(x + b) or scale them by c*g(x), d*h(x), or some combination thereof.

Wow, I guess I totally forgot what I wanted out of h(x). Still thinking about this...

Oh, maybe something cheap like h(x) = -1/(x+1) + 1

Hi ragnar (lodbrok?)

I tried to upload an image to clarify what I'm asking but the uploader wont have any of it. I'll try and clarify. I have data set that is modelled as exponential upto a transition point, lets say x_trans, after which I would like to model it as a linear function. My problem is that I dont want to introduce a dicontinuity at x_trans.

So my piecewise funcion (in my head at least) looks like this,

If x < x_trans then y = e^(x)

If x > x_trans then y = mx + c

Does this clarify the situation?