| Author |
 Topic  |
|
|
qwertqqq
12 Posts |
 Posted - 01/09/2009 : 1:45:39 PM
|
Origin Ver. 8 and SR1, Operating System: Windows XP
Hello,
I need to find the autocorrelation function of the data provided in a single column in the worksheet. I am using Analysis/Correlate dialog and then choose this column as both the input data 1 and data 2. The result however, does not seems to be what I have expected. Is there any known issue with the Origin correlation function, and is there any way to use a simple autocorrelation function?
Thank you |
|
|
VincentLiu
China
Posts |
Posted - 01/12/2009 : 04:34:09 AM
|
Hi Bob,
It seems the operation you've done is right. To find what has happened about this function, could you tell us which options you have selected in the Correlate dialog (A screenshot of this dialog is better.). And could you send your data through the following link or directly paste the data here.
http://www.originlab.com/www/company/qform.aspx?s=1&
And please tell us what kind of results do you expect?
Best regards,
Vincent
OriginLab Tech Supports
|
 |
|
|
qwertqqq
12 Posts |
Posted - 01/21/2009 : 06:56:10 AM
|
| The files are too large for upload (~300Mb), so I will describe without upload. The function is measured at several points, the point positions are collected in the X column. The function output for these points is collected in the Y column. The points in the X column go from 1e6 to 1e8 with an interval of 10. In the Correlate dialog, I select the Y column for autocorrelation. Sampling interval AUTO. Options: type LINEAR, normilize YES. Recalculate MANUAL. My Y column is both input 1 and input 2. Then a line at the bottom of the dialog appears: "time is not spaced equally. The mean increment is set as interval". I do not understand, why "time" is not set equally, because in my X column the difference between each two points is 10. Then I press OK, and the output is not what I expected in a sence that the autocorrelation does not go to zero at reasonably large X. |
Edited by - qwertqqq on 01/21/2009 06:58:33 AM |
|
|
|
qwertqqq
12 Posts |
Posted - 01/23/2009 : 11:04:49 AM
|
so this problem is still not resolved
The normalized autocorrelation starts from a vaule less then one, and it does not goe to zero assimptotically |
|
|
|
qwertqqq
12 Posts |
Posted - 01/23/2009 : 11:19:04 AM
|
I guess the problem in the Linear/Circular parameter.
Can anyone provide an exact formula, showing the difference between the linear and circular calculations used in the corr1 function? |
|
|
|
qwertqqq
12 Posts |
Posted - 01/25/2009 : 11:02:58 AM
|
OK, I am trying other parameters in this meny, and have discovered that the parameter "Sampling interval" greatly affects results of the calculations, but it should not. Any ideas?
|
|
|
|
qwertqqq
12 Posts |
Posted - 01/25/2009 : 1:38:59 PM
|
| I have tried to calculate an autocorrelation function for a column filled by random numbers. The autocorrelation function does not start from 1. Why? |
|
|
|
easwar
USA
1964 Posts |
Posted - 01/27/2009 : 2:25:59 PM
|
Hello,
Circular correlation assumes the signal repeats itself in time, so the peak correlation value of 1 (if normalized) appears at both ends of the output (when i try with random data and auto correlation). Linear correlation assumes the signal does not repeat.
The sampling interval does not affect the computation of the correlation output Y values. All it does is to determine what X values to assign to the output, so that when the output is plotted, it is same scale as the x for the input...thus one can see the correlation in the same time scale as the input.
When I try with random numbers in Y, and assign the same Y data as the input signal1 and signal2 to the tool, and I select Linear correlation, and I check the Normalize check box, I get the output to be peaked at value of 1 at x=0 and it falls off to zero at larger x on both sides.
If you see discrepancies, please send your (simper) OPJ to tech support so they can check. If your opj is very large, please contact tech support and they can provide FTP instructions for you to upload your OPJ as necessary.
Easwar
OriginLab |
|
|
|
origin_283
United Kingdom
1 Posts |
Posted - 05/17/2013 : 6:58:36 PM
|
Hello,
when I calculate the autocorrelation function of one random signal, I can two column, I know one column is the correlation coefficient rho(tau). however, I do not know what is the another column. Which quantity does it stand for? In my case, I need to plot out the figures (the vertical axis is the normalized autocorrelation coefficients while the horizontal is tau).
Thank you very much !
|
|
|
| |
Topic |
|
|
|
Autocorrelation function