When is a plateau not really a plateau?

kusterer

I have read many threads being anxious over plateaus that have lasted a few weeks or a month.

Last week my scale said I had lost 6 pounds. Yeah, right. I'll probably be on a "plateau" for weeks.

Let me talk about random measurement error, and its impact on us MFPers. I am a retired professor who taught this stuff and applied it in my research. I will be as brief as I can.

1. You are not walking around weighing or measuring yourself continuously all week and taking an average. You are taking a one-time sample of the week's weight. The sample will vary from your real weight because of measurement error, maybe by a lot.

2. Scale error. Platform balance scales are fairly accurate, spring scales on our floors at home less so. They are designed to be accurate to 1-2% If you weigh three hundred pounds like me, that's a range of plus or minus 3 or 6 lbs. 297-303 @ 1%, 294-306 @ 2%. Scales are designed to weigh people standing straight up, weight balanced between both feet and between heel and balls of feet. If your read-out is down on the scale, like mine, you might have to stand in the question-mark position like I do (knees flexed, butt pooched back, shoulders and neck bent forward). Otherwise, when I look down I see something I don't like to see, but it is not a number on the scale. The question-mark position lowers the scale's accuracy. Spring scales are also less accurate at each end of their range, so a 0-300 scale is less accurate with a 30-lb kid or a 280-lb lots-to-love.

3. Body error. Our water retention varies according to how much water, caffeine, salt and a lot of other things we take in. And also, for the majority of MFPers, by the time of the month. Former food races through our GI tract sometimes, and strolls down other times. Even though I am sure we almost all follow the same "flush first, then weigh" procedure (Who wants to weigh anything that's on its way out soon, anyway.), there is still variation.

4. Measuring tape error. It is hard to get the tape at exactly the same place every times, always exactly parallel to the floor. We have to pull it a little tight, anyway, or it will fall down. Sometimes we pull it a little tighter than others.

5. The errors may cumulate, causing an extreme "outlier," like my 6-lb loss. Or they may cancel each other out and come up with an exactly correct reading. We can use inferential statistics, like they do in opinion poils, and calculate the "margin of error," the range that we can be 95% sure the weight or measurement probably actually is.

6. Or we can do something easy, like they used to do in science labs before computers and still do on Wall Street. Take a "moving average" instead of a one-time-reading. Total the last 5 (or 10, or whatever) readings, divide by five, and get a trend line that smooths out a lot of the ups and down that occur from week to week because of variation in the error. Every week, drop the earliest reading, add in the latest, and recalculate.

7. In sum, a very high or a very low reading is much more likely to be a sample error that an actual gain or loss. Unless the long-term trend line says so, you don't have to adjust anything or fret about a plateau. Just keep on keeping on.

Machafin

Nice post, very informative and has some great ideas.