How to get the rgression line tangent to the curve
1 view (last 30 days)
Show older comments
I have a sample code which allows us to compute the correlation dimension as follows.
function [obj] = correldim(X)
N = numel(X);
arrayDist = zeros(N, N);
for i = 1:N
x1 = X(i);
for j = 1:N
x2 = X(j);
arrayDist(i,j) = abs(x1-x2);
end
end
Nr = 20;
r = linspace(0.0005, 0.015, Nr);
for k = 1:numel(r);
countk = find(arrayDist >= r(k));
cr(k) = numel(countk);
end
logr = log10(r);
logcd = log10(cr);
hold on
plot(logr, logcd)
[dimc, se]= polyfit(logr, logcd, 1);
f = polyval(dimc,logr);
plot(logr, f)
hold off
xlabel('log\epsilon');
ylabel('logC(\epsilon)');obj.cd = dimc(1);
box on
end
I noticed that while plotting the regression line using ployfit with order 1, the line crosses the curve. It looks to me the coefficient of regression line is not feasible. Could please help me how to do in exact way. I have attached the figure and code for your reference.
![](https://www.mathworks.com/matlabcentral/answers/uploaded_files/150923/image.png)
Thanks
Mahesh
2 Comments
Image Analyst
on 6 Jun 2015
You forgot to give us x. Make it easy for us to run your code and help you, not hard.
Accepted Answer
Star Strider
on 6 Jun 2015
If I remember correctly (and from the Wikipedia article on Correlation dimension), you don’t need to plot the log-log regression line.
You are doing the regression to estimate the correlation dimension, and the correlation dimension is the slope of the log-log regression, or dimc(1).
2 Comments
More Answers (0)
See Also
Categories
Find more on Descriptive Statistics in Help Center and File Exchange
Community Treasure Hunt
Find the treasures in MATLAB Central and discover how the community can help you!
Start Hunting!