I'm keeping a sort of "health log" which contains my blood glucose readings, weight, and other information as I add it. It's basically an excel spreadsheet, but I'm looking at changing the format a bit to make the information more accessible.
The actual numbers are less important than the trends, at least for now. I love Excel's graphing feature, which gives me a handy visual of the trends. However, I'll probably want something more sophisticated as I start to gather more data, so I'll probably end up writing either some number-crunching applets or learn how to write Excel script macros.
I just started out with recording blood glucose levels. I originally started four times a day, but sometimes there were fewer. I'm down to about twice a day now, once in the morning (waking) and once after I get home from work before dinner.
There is a consistency to the numbers: the later numbers are always lower, which makes sense because the medication is more effective during the day since I take it in the morning.
I do the daily weight measurement in the morning as well to try and get a consistent sample time.
I've been just recording all of this linearly, but I think that I'm going to change it to row-daily, which means that each day will have a variable number of blood glucose readings. However, I think it will make the analysis of the numbers with the graphing tools a lot easier.
After about a week of weigh-ins, I have a reasonably good idea of what my average weight for the week is, as well as the noise level. And I'm really surprised: the noise level is almost 5 lbs. which works out to something like a 36dB S/N ratio. That's abysmal, really. However, that's also affected by things like how I stand on the scale, temperature, humidity, and so on (aside from my actual weight fluctuations).
What it means is that I need to do a relatively large-order decimation for the data to show accurate trends. For instance, the blood glucose track is usually a 14-day, 30-day, and 45-day track, but it's more of a windowed average (because it's simpler). Weight is very often taken on a weekly basis, but for trend analysis, an 8-week figure is pretty common.
Decimation is a relatively common tool used in signal processing. Your CD player uses it all the time-- it's the digital filter that makes your CDs sound good. It's tied to oversampling, and is primarily used as a low-pass filter to get rid of high-frequency noise.
I think I talked about this once before. I'll spare the details. The idea is that I can do weekly averages, but the data allows me to shift from a Sunday-to-Saturday week to a Wednesday-to-Tuesday week at will, and to shuffle the window around as needed. I can also make it larger or smaller as needed to filter out fluctuations as I start to see the data.
Wow, that was boring.