I am analyzing the frequency fluctuation in the song of cicada, and my data show that there are certain periodic patterns produced within the song, each pattern could be due to a certain acoustic producing apparatus or elicit distinct female response. But before confirming that hypothesis, I need to objectively separate those periodic patterns. Beforehand, I used package strucchange in R program, but it failed to segment the sequence into discrete patterns according to shape. I wonder if there are any other methods that could help me identify patterns or shapes of frequency fluctuation?
Partly a similar issue we investigated in the area of RNA sequencing patterns over the genome: http://nar.oxfordjournals.org/content/40/9/e63.short
The shape comparison functions are in the R package rnaSeqMap in Bioconductor http://bioconductor.org/packages/2.12/bioc/html/rnaSeqMap.html, also an implementation of double-sliding window algorithm for finding significant regions. Maybe this will give you some food for thought :)
I think you need to do the Discrete Fourier Transform of your data. In this way, you will find the number of mesures (the period) each event is repeated periodically, even you don't know them. You will obtain a simmetrical "spectrum" so you will have to interpret only one half. In this "spectrum", each peak corresponds to an event that is repeated periodicallty. I have done using Mathematica software (from wolfram Research), although I suppose R or other software can do it too. If you have any doubt, please contact me.
Dear Michal Okoniewski, many thanks for your suggestion. I have searched and tried to understand the method but it is hard for me to master though.
Dear José Alabart, since I am doing self-study about Time Series Analysis, I would pay more attention to Discrete Fourier Transform, it looks like the method to be a suitable one but I cannot assure until I do the actual analysis. If I have further question, I would contact you. Thanks!
Dear Hoa, may be you will now it, I forgot to tell that the frequency of each event is easily calculated from its position of the corresponding peak in the spectrum.
Good luck,
Hi José, could you suggest any materials about Discrete Fourier Transform in R? I could not find any tutorial that could help.
Dear Hoa, It was only a thought of mine, I don't really know. Nevertheless, this article can be helpful for you:
Airborne Particulate Matter and Mortality: Timescale Effects in Four US Cities
Francesca Dominici, Aidan McDermott, Scott L. Zeger, and Jonathan M. Samet. Am J Epidemiol 2003;157:1055–1065., that is free available online at:
http://aje.oxfordjournals.org/content/157/12/1055.full.pdf
Inside, it is call that: "SAS, S-Plus, and R software for decomposing a time
series into a desired set of frequency components can be
downloaded at http://www.ihapss.jhsph.edu/software/fd/software_fd.htm."
Best regards,
Dear José, I have read the papers you suggested, however, I do not think the method is adequate to my data. Would you mind having a look at my data? I can send you by email.
Dear José, I do not see your email address so I post the file here. There are two variables: PF stands for Peak Frequency, and PA Peak Amplitude. I would like to analyze the fluctuation of PF and PA simultaneously. Thank you for your help.
Hi, Hoa. Excuse me for the dalay in aswering, I have been very busy.
In the sample you provide me (7430 points) I have identified the main peaks in the DFT in simultaneous pulse frequency (PF) and pulse amplitude (PA) at positions 5, 55 and 58, corresponding to period values of 7430/(5-1), 7430/(54-1), 7430/(57-1), that is, 1865, 138 and 130, that is, something is repeated (at least twice in the sample) every 1865, 138 and 130 points, If sample points were taken every i.e., 1 miliseconds (ms), something was repated every 1865, 138 and 130 ms. To see what it was repeated is not so easy, although I have identified these putative repetitions.
I identified other more frequent repetitions (every 3 or 4 points), I discarded them as I considered them not relevant.
Have all of this sense for you? Please provide me your mail or write to me ([email protected]) if you wnat more details. I can send you these findings.
Best regards,
Thank you so much for your help. Please give me sometime to understand what you have written.
- Badges
- Science topic
- Similar topics
- Medicine
- Physiology
- Biosignals
- Biomedical Signal Processing
- Statistical Analysis