Libra Weight Manager (App Settings)

Options
leggup
leggup Posts: 2,942 Member
I just started using Libra and I love it. For those who don't know, it's a weight-trend app. Put in a bunch of data about your weight and it creates a smoother trend line, including predictions. My question for users of Libra specifically- there are two settings I'm not sure about:
  • Smoothing days: Controls how strongly each weight value affects the trend
  • Forecast days: Controls how many days to use for the forecast calculation
Both settings started at the default of 7 days. I weigh in daily, so each week has 7 points of data. Would you use 7 full days of data for the forecast? More? Less? For smoothing--- I don't have much outlier data-- I'm assuming a setting of 2 means that the smoothing is low, while 7+ would be high. If you use another weight trend app, like happy scale, how much smoothing does it do? How many data points for a forecast?

Replies

  • AnnPT77
    AnnPT77 Posts: 32,055 Member
    Options
    Play with them. You can always reset them back. The settings don't affect how it stores the data, just how it builds the displays.

    As a Libra user for around 2 years, here's how I understand them, in nontechnical terms:

    Loosely, the smoothing days affect how spiky/volatile the trend line is. Assuming your actual trend is mostly in some general direction - like you're generally losing over a period of weeks/months - more smoothing days will make the trend line less jumpy/pointy as it goes in that general direction - a little less perturbed by daily fluctuations. More smoothing days means it's considering a longer set of days when it builds the trend line, so single individual up or down days "pull" on the trend line less.

    For example, I'm currently losing super slowly in maintenance to drop back down in my maintenance range - probably averaging a pound a month or less. A smoothing days setting of a couple of months will show this slooow trend, while 7 days just shows a peaks and valleys trend line. While I was actively losing, the same 7 day setting showed a somewhat spiky but visibly downward trend..

    Forecast days is about how it predicts your future weight loss. If you're losing. at a consistent rate (pounds per week), changing this may not make much observable difference. If your loss is speeding up or slowing down, longer settings will make it slower to "recognize" this and change the angle of the projection line. It's about how many days' trend it uses to make the future forecast.

    This is just my experience and understanding. I know enough about statistics to know generally what's going on here, but not enough to explain in statistical terms. Hope it helps.
  • misnomer1
    misnomer1 Posts: 646 Member
    Options
    misnomer1 wrote: »
    The trend formula for Libra is

    current trend = Alpha x latest weight entered + (1 - Alpha) x previous day's trend

    where, Alpha = 1/smoothing days as input in settings

    Basically, the more the smoothing days, the less weight is given to latest data you entered and more given to previous Trend weight (not previous weight you entered). This method is 'exponential smoothing'.

    Forecast basically takes the trend weight for whatever days you have chosen and extrapolates (start trend weight and end trend weight of the period chosen) it as a straight line.
  • leggup
    leggup Posts: 2,942 Member
    Options
    That is helpful. I don't see much of a change in terms of either since I only have 7 days of data in to play with, so far. I am more interested in the prediction line since my goal is to lose 23 more lbs by March. At least right now (with my limited data) it is using a linear trend line, which is misleading. Although I can lose 1-1.2 lbs a week right now, I'm going to slow down to less than a pound per week pretty soon. The forecast says, "You'll reach your goal in Jan!" which is inspiring, but unrealistic.
  • AnnPT77
    AnnPT77 Posts: 32,055 Member
    Options
    leggup wrote: »
    That is helpful. I don't see much of a change in terms of either since I only have 7 days of data in to play with, so far. I am more interested in the prediction line since my goal is to lose 23 more lbs by March. At least right now (with my limited data) it is using a linear trend line, which is misleading. Although I can lose 1-1.2 lbs a week right now, I'm going to slow down to less than a pound per week pretty soon. The forecast says, "You'll reach your goal in Jan!" which is inspiring, but unrealistic.

    The projection line is always linear.