polyfit semilogy

Hi,

I'm trying to find the fractal dimension of a range of data, but I'm sruggling looking for a way of finding the best range of data in a bigget set of data, which will fit the best regression line in semilog.

do i have to try every possibilitie,

 A=polyfit(X(1:1000),log10(Y(1:1000),1);
 A=polyfit(X(2:1000),log10(Y(2:1000),1);
 ...
 A=polyfit(X(i:j),log10(Y(i:j)),1);

and so on.. or is there easiest way?

Thank you for your help

n.

[Andrew Newell]

What are X and Y?

[Nicolas]

for example

X= 1500:500:7000

Y= 4 15 30 59 96 151 214 291 380 480 598 741

I've tried to take out the points with the highest residuals, but it doesn't work very well. I need to truncate my data from the ends towards the middle to find the best fit, instead of taking out the middle point to make the regression line better.

Answer 1

[bym]

do you mean stepwise regression?

doc stepwise()  % assuming you have statistics toolbox

[Nicolas]

hi, do you think stepwisefit would be better in my case?

I'll work on that, both sounds interesting ways!

Answer 2

[Andrew Newell]

If X vs Y is a box count plot, you could calculate a local scaling exponent.