Averages (7, 14, 30, 60, and 90 days)

[issysmommy]

Hadn't looked at my averages lately, but just started Weight Watchers and been upping (is that a word) my activity level lately....I was happy to see the following...just want to get it even better...I think I'd like to see my averages be right at 100, but I'd settle for something around 120. As a mathematician, I think it has to be noted that these averages are probably not accurate as when my BG is low I tend to test more often, so that would throw more lows into the data. I am thinking that a median (the middle number) might be a more accurate measure of central tendency to use here, although the frequency of the low tests would again affect the data. Hmmm...

Days 7 Tests 70 Average 146
Days 14 Tests 148 Average 155
Days 30 Tests 242 Average 184
Days 60 Tests 522 Average 192
Days 90 Tests 820 Average 189

[gnstriker]

how are the numbers when you remove outliers?

[Dervish]

As a mathematician, I think it has to be noted that these averages are probably not accurate as when my BG is low I tend to test more often, so that would throw more lows into the data. I am thinking that a median (the middle number) might be a more accurate measure of central tendency to use here, although the frequency of the low tests would again affect the data.

Yeah, that was one of my first thoughts when I got my first meter. "These averages look like they treat every reading equally. They're worthless."

I think the simplest improvement would be to weight each reading based on the time between it and the two readings adjacent to it. For a more realistic average, you could switch to a two-pass process, averaging all readings within each hour in the first pass (hours with no reading could use the previous hour's average or interpolate a value, depending on how complex you want to get) and then averaging the hourly averages in the second pass. Either option would prevent the average from being unduly affected if you read a 50 and then re-test every 5 minutes until you get back up to your target range.

[Lloyd]

You can see a lot of things from a graph, that are not easy to see from the numbers.

-Lloyd