Drawing inference from WSPR spots database

38 posts / 0 new
Last post
VK1OD
Offline
Last seen: 4 years 5 months ago
Joined: 2009/06/08 - 21:30
Drawing inference from WSPR spots database

I am interested in drawing some meaningful inferences from the huge collection of data in the spots database.

I have done a quick frequency distribution on SNR on about 600,000 observations for July so far, and find that the range of most spots is from about -20 to about +10dB.

The distribution is fairly linear between those limits, increasing with decreasing SNR as might be expected.

Whilst I did expect the upper limit behaviour to be as discovered, I was somewhat surprised that the frequency of observations with SNR below -20dB falls away quite quickly. That would suggest that although the detector might have a theoretical decode limit of much lower than -20dB, other factors like perhaps impulses, minimum duration of adequate SNR, etc enter into the response. It suggests that obs with SNR below -20dB distort the picture.

Given that the distribution is so skewed, I think it is not suited to analysis using parametric statistical techniques, and resort to non-parametric techniques will be necessary.

That being the case, the matter of tx power enters into the usefulness of the data. Given the distribution, I cannot see a simple and reliable way to account for the tx power variable, other than to select only observations using the same tx power for summary. Fortunately, the data suggests that more than half the stations transmit with 5W (37dBm), so selecting them alone does not discards only half the observations whilst eliminating the tx power variable in the analysis. BTW, I did see in the sample set that tx power ranged from =40dBm to 60dBm. Makes one wonder!

Any comments?

Owen