Linear Regression on log log plots

Hello – I would like to plot seismic data with a y range from 10 -7 to 10-1 (m*s) and an x range from 1 to 10 000 (Hz) on a log log plot. Than I would like to do a linear regression on this but basic fit gives a line not reflective of the data. Could you please advise what is the best technique to use ?

Any help will be much appreciated.

Answer 1

[dpb]

X=log10(x); Y=log10(y);  % convert both variables to log's
b=polyfit(X,Y,1);        % estimate coefficients
yhat=10.^[polyval(b,[X(1) X(end)])];  % evaluate at end points
loglog(x,y)
hold on
loglog([x(1) x(end)],yhat) % add fitted line

[Rookshana Trollope]

Thank you I will try this

[Rookshana Trollope]

It says unexpected matlab error. must I ignore line 4 and 5 ?

[Rookshana Trollope]

I get as far as b but yhat gives me an unexpected matlab error.

[dpb]

oops, use the "dot" operator .^ to take element-by-element power

[Rookshana Trollope]

Ok will try - Thanks I dit that and nothings happening after 30 min. But it looks

like its still trying so it must be in a loop. Can I load an screen image for you to look at ? will really appreciate your help.

[dpb]

Note the message at the bottom -- "Continue entering statement"

It's waiting for something but I can't tell what--the paren's look to be balanced unless my old eyes are missing one.

Try another CR or two; if that doesn't finish the line then Ctrl-C out back to the command prompt and try again...either there's something in the commandline processor that's hidden or the graphics engine in a newer release has gotten hung up or somesuch; with a 2-length vector it would take only a few seconds at most.

[Thomas Schragl]

when trying the code noted above I found that there is one ] missing in line 3

right now:

yhat=10.^[polyval(b,[X(1) X(end)]);

and should be:

yhat=10.^[polyval(b,[X(1) X(end)])];

note the ] before the ;