I’d like to try different curves against data to see which of them would best fit. So in this example x0 = collect(1.0:100.0) y0 = x .^ 4.5 f0 = curve_fit(LinearFit, x0, y0) I’d like to learn just how bad it is, but is there an easy way to get that out of it?

I have a very simple package to make fits really, really simple, in common cases: GitHub - m3g/EasyFit.jl: Easy interface for obtaining fits for 2D data

Just in case it might be useful, sharing one linear regression example below, where the correlation coefficient r is displayed and other regression statistics are computed too: using Plots, Random, Distributions, GLM, DataFrames, Printf # input data n = 30 x = LinRange(0,1, n) .* 10 e = rand(Norma…

Thanks, I’ll look into these examples. The problem itself should be relatively simple, the actual data is a straight line in a log-log plot, it’s just not quite that easy to post its code here. (The wikipedia page for log-log plots is also somewhat opaque.)

Check out this document too on lin-log and log-log regressions .

Just for fun: To fit a polynom you can use """ Fit y = f(x) to nth order polynomial. Returns list of polynomial coefficients. """ function polyfit(x,y,n) @assert length(x) > n A = [k^n for k in x, n in 0:n] A\y end To reason about the fit you would need to look for measures often found under g…

[image] feanor12: A = [k^n for k in x, n in 0:n] Don’t do this — the Vandermonde matrix is generally ill-conditioned .

I got this working with minimal effort (on the package side, I had some other unrelated Julia problems though). When I take log10 of the original x and y values, and do fit = fitlinear(x_log,y_log), I get a nice match. [Screenshot 2021-04-08 at 15.09.00] I can also see that the average square res…

When I tried to fit against the original (non-log) data, the only one that gave good results was fit = fitspline(x,y). [Screenshot 2021-04-08 at 15.11.29] It’s just that spline doesn’t really tell me anything useful. -------- Spline fit --------------------------- x spline: x = [0.001, 0.011…

A spline will always look well, but you cannot use them to really get any information about your data. If your x and y values are related by something like y = A/x it is expected that the log/log plot is linear (log(y) = A - log(x)) while the plot of the original data is not.

Yes. X ^ -1 is close. Can I fit that with EasyFit?

Simple curve fitting with CurveFit (or alternative)

General Usage

rafael.guerra April 7, 2021, 10:22am 2

Please check this post.

1 Like

Topic		Replies	Views
Linear Fit New to Julia curve-fitting	8	5390	August 3, 2022
Plot fitted curves New to Julia question , curve-fitting	6	1564	March 21, 2022
Alternatives to LsqFit for nonlinear curve fitting? General Usage curve-fitting	12	3311	December 21, 2023
Exponential Curve Fit Equation New to Julia question	6	2388	March 28, 2022
How to use the output of log_fit in abline plotted in a loglog plot in Julia General Usage question , plotting	4	1288	December 23, 2019

Simple curve fitting with CurveFit (or alternative)

Related topics