On the My ADS-B page, there is a two row table of numbers for many days’ worth of data. This appears just under the circular histogram. Further down, there is a chart showing neighboring receivers and daily mean stats (parenthetically described as past week). I assumed the parenthetical remark indicated that the stats were a mean over the last week, or in other words an average of the last seven columns of the first chart I mentioned.
However, I note that the stats listed for my station are identical to the 7th column of the first chart. Is this a bug, and the second chart is showing the 7th oldest day rather than a mean over 7 days?
I thought that the description of the column is clear enough - Daily Median Stats (past week). It is the median of what was recorded for your station over the last 7 days.
I suspect the 7th column is the median of your last week, that chart shows the median, not the mean… So when you look at your last week, ordered by number of aircraft tracked and positions reported, your 4th data point should be what’s represented.
To me, it should be the mean, so you can get an average over the week, but that’s just me.
Oh, right. That’s what I get for copying/pasting instead of typing. I was reading mean when I pasted median.
However, in order for the median to calculate correctly (to match what’s displayed), I have to include the first column, which is incomplete (today’s data is still accumulating). So it’s really sort of a 6 1/2 day median.
Interesting. Now the median is numbers that don’t exist anywhere in my data as far as I can tell. They are just a tad higher than today’s totals. Since it’s just after 00:00 UTC, I suppose it’s just the servers busy doing end of day processing, especially, since I don’t see anything for 1-7 yet.
IMHO, the past week median should not include the current day since it is incomplete and thus can only skew the figures. I wondered why this shows the median and not the mean for a long while since mean is way simpler to compute (AFAIK). A searching shows that median is statistically more stable than mean when you have values that are way off. So it is preferred by people doing stats over simple average.
I did a sample spreadsheet over my own data and the median looked better over time than mean with a difference over the average that was pretty small.
Median, in this case, is simpler to calculate. Just take the 7 values, omit the highest 3 and the lowest 3. What you have left is the median. For mean, you have to add the values together and divide by the number of items.
Another option is a sort of in-betweeen measure, which first deletes the outliers and then takes the mean of what remains.
