|
Community
Home
eHam.net Home
Articles & Stories
Speakout
Strays
Survey
My Profile
Resources
This Week's Contests
Classified Ads
Contest Links
Product Reviews
Site Information
About This Site
Contesting.com Team
|
|
1-2 of 2 messages
|
Page 1 of 1
|
|
CQWW CW Two Radio Event Signatures
|
Reply
|
|
by zs6aaa on February 20, 2008
|
Mail this to a friend!
|
This is a very interesting study, and I appreciate the effort that the authors have made to devise a measure that may be used to improve the integrity of contest results by highlighting entries that may be in the wrong class.
However I question whether the statistical measure they chose is the correct one. The measure they use is defined as follows
"We call the 2RE probability distribution for a station or class its two radio event signature. The probabilities were computed by dividing the number of 2REs at each run rate by the total number of QSOs logged." (p. 3)
However because they divide by the total number of QSOs made (at all run rates) rather than by the number of QSOs made AT THE PARTICULAR RUN RATE IN QUESTION, this measure will be biased depending on the statistical distribution of different run rates.
For example, consider these two fine contesters
Contester A
Rate Hours QSOs TREs prob.
1 16.0 960 96 0.028
2 8.0 960 24 0.007
3 4.0 720 8 0.002
4 2.0 480 3 0.001
5 1.0 300 1 0.000
Total 31.0 3420 132
Contester B
Rate Hours QSOs TREs prob.
1 3.8 228 23 0.007
2 3.8 456 11 0.003
3 3.8 684 8 0.002
4 3.8 912 6 0.002
5 3.8 1140 5 0.001
Total 19.0 3420 52
(Sorry, eham.net messes up the table by concatenating adjacent spaces).
Contester A works for 30 hours and gets 3420 QSOs. He has 16 hours at a rate of 1 QSO per minute, with exponentially fewer hours at higher rates. His number of second-radio QSOs is an arbitary decreasing function of rate which I chose for this illustration to be
TRE = 0.1 * QSOs / Rate^2
Contester B works for 19 hours and also manages 3420 QSOs. However she works 3.8 hours at each of the rates (1-5 QSOs per minute). Her chance of a second radio QSO is calculated using exactly the same function as for contester A.
Now have a look at the "2RE probability distribution" for the two contesters (last column of table). contester A has a high peak at Rate 1, and the tail drops off rapidly at higher rates. Contester B has a lower peak at Rate 1, but a broader "right hand tail".
These characteristics (a lower peak 2RE probability distribution and a broader "right hand tail") are precisely the factors quoted as being suggestive of "assisted" operation (either by packet, or as a M/2 station masquerading as SO2R). The assumption is that "assistance" of some form results in a greater ability to make second-radio QSOs at high rates, and this is the cause of the broader right hand tail and lower peak 2RE probability.
However this is incorrect, as Contester B's second-radio QSO rate was calculated using exactly the same function as for Contester A. The reason is much simpler: she worked at high rates for longer than contester A, which is why she had more second-radio QSOs at high rates. This translated into a higher "2RE probability" because the probanilities for A and B were both calculated by dividing the number of TREs by the same TOTAL number of QSOs.
The danger is that different statistical distributions of run rates (perhaps due to differing propagation, or different band strategies) can skew the proposed measure of "2RE probability distribution" in a way that is hard to distinguish from the signatures associated with assisted operation.
The problem can be solved by redefining the "2RE probability distribution" as the number of second radio events at some rate R divided by the number of QSOs made at that rate R.
Andrew ZS6AA
|
| |
|
RE: CQWW CW Two Radio Event Signatures
|
Reply
|
|
by VE5ZX on February 21, 2008
|
Mail this to a friend!
|
Andrew,
Thank you for your interesting and insightful comments.
When we started the research project more than a year ago we did the analysis using the method you suggested. I have placed an Excel workbook on the Radio Sport Canada server containing Figure 5 from the article re-done using the technique you suggested. It is available from the URL below
http://www.radiosport.ca/rsrp/documents/Modified%202RE%20analysis.xls
As you can see both analytical technique show that the more assistance an operator receives from packet or another operator the more likely a 2RE will occur at higher Q rates. We chose to use the method we did in the article because it is a more intuitive method of presenting the findings. Both representations can be used to provide supplementary information to adjudicators.
The constant time example that you gave is unrealistic and artificially creates the broadening in the right hand tail. Even the most skilled operator has no control over the amount of time he/she can spent at a given Q rate. In your example contestant B worked 3.8 hours or 228 minutes at a Q rate of 5Q/minute. In the 2006 CQWW CW contest only PJ4A worked over that amount of time (305 mins) at 5Q/minute. All other contestants in the SOAB, SOAB A and MS classes spent well below that amount of time at that Q rate. The average time spent at 5 Q/minute was about 43 minutes.
Out of curiosity I prepared a graph that shows the distribution of the fraction of the total operating time an average station in each subclass spent at each Q rate. It is in the Excel workbook referred to earlier. You can see all subclasses exhibited a rapid decay of time spent at various Q rates in the right hand tail of the time distribution. Less than 5% of the total time was spent at 5Q/minute. The broadening effect you discussed is much less pronounced in the actual data compared to the example you gave.
Thank you for your comments
Radio Sport Canada
Syl, VE5ZX
|
| |
|
Email Subscription
You are not subscribed to this topic.
Subscribe!
My Subscriptions
Subscriptions Help
Check our help page for help using
Forums, or send questions, comments, or suggestions to the
Forums Manager.
|
|
