Forum: help
Monitor Forum | 
Start New Thread
| RE: CalculatePowerBand settings [ Reply ] By: Stefanie G on 2020-02-07 16:59 | [forum:47371] |
|
Can you provide more information about the appropriate size of window for fourier analyses? For example, we have 5 minute recordings and want to examine HF HRV for 1 minute epochs. We would use the SplitPowerBandByEpisodes to examine HF HRV for these 1 minute epochs. What window size is appropriate? If the minimal HF frequency is .15, then our window size should be greater than 6.67 (1 divided by .15). Our question is whether 7 seconds is too small of a window for 1 minute epochs in a 5 minute recording. Or should we just do 60 second window? Thanks in advance!! |
|
| RE: CalculatePowerBand settings [ Reply ] By: Constantino Antonio on 2018-06-26 16:10 | [forum:46084] |
|
Dear Harry, Sorry for the late reply. When analyzing any time series by windows, the size of the window is limiting the minimum frequency that you can resolve (Think about it, intuitively, you can't see what happens in a time scale of years by looking at it in windows of 5 seconds). As a rule of thumb, to resolve a frequency fmin, you should use a window size of the order W > 1 / fmin. Concerning the wavelets, they surely can be used in short recordings. In some sense, the multiresolution analysis that can be derived from the wavelets theory picks the proper window for you: it uses large windows for the low frequencies and small windows for the high ones. Hope this helps Kind regards |
|
| CalculatePowerBand settings [ Reply ] By: Harry Fisher on 2018-05-24 08:24 | [forum:46018] |
|
Hello! I am wondering if there are any guidelines for setting the "size" and "shift" arguments for the CalculatePowerBand function using the fourier method. I have 5 minute baseline RR recordings and would like to analyse the HF component, but I'm unsure of the appropriate settings for such short recordings. I've tried setting size = 60, shift = 30 and then taking the mean of the values calculated, which seems to give a sensible result, but wondered if there might be a better way... Additionally, would there be any value in using the wavelet method for short recordings? Many thanks, Harry |
|
