Hello,
I have some data that looks like the line on the right, but I'm not sure what function to use. It seems like an exponential added to a line, or something of the like. I think it may be the same idea as what I have on the left, which is a hyperbola added to a line. Any ideas?
Thanks!
Nathan -
If you want to use two equations, one before and one after a certain x-value, then you need to specify it that way. (This before x=whatever, and this other equation for larger x. You write that by stacking one equation over another and putting the condition for x next to each, using ,and = signs.)
Are you talking about experimental y values for which you have one predictor, x? Maybe one quadratic or some other continuous function fits. This would mean using regression and a graphical residual analysis to see if your model fit is good (and a cross-validation to see if you overfit to your points - perhaps just hold out some data and see how well your model "predicts" those y-values). If you are only having measurement error, because this is some known relationship from physical science, this might help you find the best curve, but if it is a matter of modeling something not known, then you have more of a standard regression model problem.
You might need to explain more about what you are doing.
There is an overlap of your curve and straight line that is confusing.
Best wishes - Jim
Updated curve posted in this reply.
This is for a fluorescein-labelled RNA oligo standard curve. It seems at low concentrations the signal loses linearity, perhaps due to sensitivity problems. I did subtract the background.
I fit the the data with a second order polynomial yielding a great fit with r2 value of 0.9999, so that equation will definitely suffice for converting RFU to product concentration. However, I can tell that the polynomial curve never becomes completely linear, as my data seems to, and I want to hone my data fitting skills, so I am trying to find a better fitting model.
Also, I'm not sure that it requires 2 equations. I think you can add curves together in the same equation. Like the hyperbola one I showed. I think that is appropriate, because I think the factor that is causing there to be a linear element is present in the early part of the curve, it is just added to the shape of the hyperbola. Not positive that's an appropriate assumption though.
What about a sigmoid function, 1/(1+exp(-x))?
https://www.wolframalpha.com/input/?i=1%2F%281%2Bexp%28-x%29%29+for+x+%3D+-5+to+0
Regards,
Joachim
- Badges
- Science topic
- Similar topics
- Cognitive Science
- Thinking