Login

Image

Trendline

Ideas and Improvements

Can you ad a trendline into the graph, that you can see if your on your way to reach the goal.
Simply a doshed line from startdate an weight to goaldate and weight.
about 5 years ago by Anonymous
I kinda like how trend line is done here -> http://physicsdiet.com/Charts.aspx?t=WeightLoss

about 5 years ago by jenni
The tool I use looks like this:

Red = Trendline
Yellow = Goalline
about 5 years ago by Anonymous
I'll have a look into this soon. Not sure about the math (not my strong point) ;)
about 5 years ago by D.Fox
The math is kinda simple - the trendline follows the average weight. Say you weight following numbers during a week - 109.3, 108.9, 107.9, 108.1, 108.0, 107.7, 108.0.

So on first day the trendline would be 109.3 since that when it starts. The second day the trendline would be (109.3+108.9)/2 = 109.1 and the trendline would be going down because you weighted less (108.9) than the average weight (109.1). Average weights for the following days would be 108.7, 108.5, 108.44, 108.31 and 108.27 - so the trendline would be going down all the time.

The idea of the trendline is to motivate by showing that your weight is going down on average, even if single measurements bounce back and forth. Normally you might think you've failed when your weight goes from 107,7 to 108,0 (in the example), but the trendline shows that since you're weighting yourself underneath the average weight, you are indeed losing weight. Only if you start weighting above the average weight, will the trendline start to rise as well. In the beginning it takes a while for the trendline to actually "catch up" with your current weight.

Hope this made some sense :)
about 5 years ago by jenni
I've added a trend line based on the suggestions. Looks nice so far :-)
about 5 years ago by D.Fox
You can also use a smoothed moving average. Than the new value has a bigger weighting than oder older ones.
Ad = Average of the day
Ad-1 = Average of the day before
Md = Measurement of the day

Ad = Ad-1 + 0.1 * ( Md - Ad-1)

The 0.1 is the smoothing percentage in this case 10%, what works fine.
about 5 years ago by Anonymous
With the same data as jenni it will look like this.
The trend ist the difference between weight and average.

for 10% (0.1)
weight	average	trend
109,3 109,3 0,0
108,9 109,3 -0,4
107,9 109,1 -1,2
108,1 109,0 -0,9
108,0 108,9 -0,9
107,7 108,8 -1,1
108,0 108,7 -0,7


and 20% (0.2)
weight	average	trend
109,3 109,3 0,0
108,9 109,2 -0,3
107,9 109,0 -1,1
108,1 108,8 -0,7
108,0 108,6 -0,6
107,7 108,4 -0,7
108,0 108,4 -0,4

about 5 years ago by Anonymous
Should the time between the measurements be taking into account as well?
about 5 years ago by D.Fox
D.Fox: The website that Dino referred puts more value on recent entries (within week or two weeks) than ones that are over month or more old, so I've understood, to keep up with the current development.
about 5 years ago by jenni
Im using the book "The Hacker's Diet" by John Walker
You can get it from here for free.
The Book described in one chapter how to visualice your weight progress. There are examples why to do that and how to calculate all this stuff like moving averages.

I don't know how it works if you work with the time between measurements, it's more like a daily collecting date, because if not the average of your weight stagnates. Like here when people only fill in there data every few days the average looks like only something happend the day before the new data.
So I'm not really sure how it would work.
about 5 years ago by Anonymous
Dino: I'll have a look at that formula over the weekend. It's a bit more difficult to implement that the thing we have right now.
about 5 years ago by D.Fox