Posted on Tuesday, 17th April 2012

Please read Section 18.3 and post a comment.

Posted in Class | Comments (11)

  1. Shubham Debnath Says:
    April 18th, 2012 at 10:24 pm

    Can you use a Gaussian filter instead of linear filter?

  2. Ben Dichter Says:
    April 19th, 2012 at 12:01 am

    Is the Nyquist frequency a hard boundary? What if there is noise in the data?

  3. Eric VanEpps Says:
    April 20th, 2012 at 5:18 pm

    In section 18.3.3, you discuss 5 different smoothing methods. Are there any good reasons for choosing one method over another, beyond convenience? Do any of the methods have slightly more desirable properties or are any of the methods more accepted as normal?

  4. Rob Rasmussen Says:
    April 23rd, 2012 at 9:39 pm

    What is the difference in effect between smoothing the periodagram versus detrending the data prior to estimating the periodagram? It seems like they might be somewhat similar, but it is hard to tell.

  5. Jay Scott Says:
    April 24th, 2012 at 1:39 am

    p538

  6. Jay Scott Says:
    April 24th, 2012 at 1:41 am

    p538 on the introduction of bias: Is there a way to gauge the magnitude of the bias relative to the variation in the log periodogram. Is this an intuitive judgement call or one for which rules exist?

  7. Rex Tien Says:
    April 24th, 2012 at 6:04 am

    Are there any kind of tests to determine if aliasing is going on in your data? Perhaps by varying sampling frequency methodically?

  8. David Zhou Says:
    April 24th, 2012 at 6:38 am

    I really liked the detail that you go into about how to create a spectrogram, thank you.

  9. Sharlene Flesher Says:
    April 24th, 2012 at 8:01 am

    In 18.3.3, it is emphasized that smoothing of the periodogram is critical to get a good estimation of the spectral density function- how much of the data gets lost in the smoothing processes? Of those mentioned, are some more or less conservative about smoothing over important fluctuations?

  10. Kelly Says:
    April 24th, 2012 at 8:21 am

    In linear filtering, I’m not quite sure why it’s necessary to pad the Fourier transform with zeroes if the filtered range is really what we’re interested in.

  11. Thomas Kraynak Says:
    April 24th, 2012 at 8:28 am

    Is there a formal procedure that can be used to determine the time intervals for computing spectral density estimates, or is it mostly arbitrary/intuitive?

Leave a Reply

You must be logged in to post a comment.