Originally Posted by kv501
Less than .5 lbs?
Since this is the Clyde forum and just for arguments sake let's say I weigh 250 lbs. .5 lbs is two tenths of a percent of my body weight.
I think you're not going to gain anything but frustration trying to do that; your body weight will vary as much as 3-4 lbs per day depending on what you've eaten, when you've eaten, time of day, and a whole host of things. Just weigh yourself once a week if that. There are way too many outside factors to be breaking your weight loss down to less than 6 oz. increments.
There are at least two distinct schools of thought on the precision/frequency of measuring weight question. There is certainly a lot of daily variation connected with the factors you've described, as well as with things like how much water your body is holding onto that day, how much sodium you took in the last few days, etc.
This "noise" can be modeled as a random process added to something else we can call the "real" weight, which will vary very little from day to day. Because the noise swamps the "signal", which in this case could be a trend by which the actual weight changes by as little as a tenth of a pound per day, we usually try to integrate out the noise. The assumption is that the signal is highly correlated from sample to sample, but the noise is close to uncorrelated from sample to sample, so the standard deviation of the average of N samples decreases as the square root of N.
If you weigh yourself once a day, trying to control for extraneous factors, i.e. weighing yourself at the same time every day, without clothing, etc, and then average your measurements for the week, the standard deviation of the noise would be reduced by a factor of 0.38. So, if the original noise variance was 1 lb, the standard deviation of the average of your measurements would be 0.38 lbs, making it easy to see a real weight change of as little as half a pound with low probability of error.
Again, assume the variance of the noise is 1 lb. Imagine - if I weigh myself in one week, and then weigh myself a week later and weigh 2 lbs more than I did the last time, was it because I gained 2 lbs? Or did I happen to hit a "low" day the first time, and a "high" day the next. Eventually, after months, I'd be able to see the trend. But it would take months.
This isn't so important if you're sticking to a program religiously, and losing quickly - tracking and detecting changes serves more the purpose of reinforcement than anything else. But this time delay in detecting a true change can really kill you if you're on maintenance, and your real weight isn't supposed to be changing quickly. It could take you months to detect a trend of +1 lb/month - which, if not dealt with, would result in gaining more than 10 lbs in a year.
Percentage of body weight is irrelevant - whether you're tracking differences on the order of 0.1 lbs per day in "real" weight on a 250 lb body or a 100 lb body, the noise still has a variance of between 1 and 2 lbs, which totally swamps the signal - the signal-to-noise ratio is on the order of -10 dB or worse.
So I'm in firmly in the weigh every day camp, but smoothing the measurements by averaging. The problem with a wonky scale is that, in addition to the noise that my body adds to the measurements, the scale is adding another, uncorrelated noise process ... which stinks.
I've also noticed (and this is just for laughs), a periodic component in the noise, with a period of about a week. I see this when I look at the graphs, but it's hard to pull it out because of the quantization error introduced by the coarse granularity of the scale. That's why I wanted better than 0.5 lbs precision - to reduce this so that I could see if there's really a 7-day oscillation. Again - this has nothing to do with losing or maintaining. It's just curiosity. For what I really need to do, 0.5 lbs precision is fine.