How do weight trending apps decide your trend?
purple4sure05
Posts: 287 Member
Hey guys. I've been using libra for a couple of months now every day as I've worked on transitioning into maintenance. For the first month even when I had high weigh ins, it insisted my trend was continuing downward.
About a week ago I had a high carb, high cal day that I banked calories for. My weight went up for a few days in a row,less than two pounds from bottom to top (130 to 132) . For a few days now I've been back down closer to 130 and my app still insists that I will gain 3/4 of a pound a month based on that gain after my high calorie day (2200)
I only ask because I'm curious how this app is meant to actually help me? For the last 5 or 6 weeks my weight has stayed between 129.7 and just over 132 up and down, so I dont see how the trend suddenly spiked upward. Do these apps not accurately predict trends, or is there a good reason why it still seems to think I'm destined to gain significant amounts of weight? I'm just asking here because I'm curious how to properly use this app to help me understand my weight trends. I dont want to stress over an upward trend if its unfounded.
About a week ago I had a high carb, high cal day that I banked calories for. My weight went up for a few days in a row,less than two pounds from bottom to top (130 to 132) . For a few days now I've been back down closer to 130 and my app still insists that I will gain 3/4 of a pound a month based on that gain after my high calorie day (2200)
I only ask because I'm curious how this app is meant to actually help me? For the last 5 or 6 weeks my weight has stayed between 129.7 and just over 132 up and down, so I dont see how the trend suddenly spiked upward. Do these apps not accurately predict trends, or is there a good reason why it still seems to think I'm destined to gain significant amounts of weight? I'm just asking here because I'm curious how to properly use this app to help me understand my weight trends. I dont want to stress over an upward trend if its unfounded.
1
Replies
-
Can you share a screen shot of your trend screen? May help us understand better.0
-
They don't work for me so I gave up on them.0
-
0
-
I don't use a weight-trending app, but based on the projected trend line, it looks as though it's set to calculate the trend only looking back at a couple of weeks' data. Check the settings and see if that's something you can adjust.0
-
purple4sure05 wrote: »
To me that is just what I have come to expect from it. I don't know if that's good or bad. To me, it just is. But you had 5 days significantly over trend. You've had 3 days below it since. So your trend is coming back down, but the 5 days of being over still weigh heavy in the data set. That's why even though your trend is moving back down, you are still overall projected for a slight gain (you may have classified it as major but really its about as small as a gain can be documented while existing). I am sure there is an equation behind it all. Someone may know it better than me. But it seems from a logic standpoint that if you were up for 5 days but down for 3, that the trend would still be more up than down. If you have a few more below trend days, it wl likely move downward.
Ultimately you will decide if the app has value or not. I use it and find the data points interesting, but I don't get super caught up in the details of it.4 -
It's just doing statistical projections based on averaging over multiple days. You can change the number of days it uses.
Click on the 3 dots in the upper right, then click settings, then advanced preferences.
Speaking very loosely: The smoothing days setting affects the backwards look, i.e., how strongly individual fluctuations affect the trend line. Setting it longer (more days) will tend to make the past part of the line less bumpy. Forecast days affects how much of recent history gets used to guess at the future.
You can play with these settings, and see how it affects the graph. Nothing you change here affects the underlying data you've recorded, so play to your heart's content, in order to understand what the settings do. If you reset back to the original values, the chart will look just the same as it originally did, so you aren't hurting anything by messing with these settings temporarily.
I was fine with the default settings while losing. (I do pretty much understand what they do, so I don't freak out about transitory weirdness.) Now, in maintenance, I prefer using longer settings (30/60 repectively) because it works better for me with uneven eating I do.11 -
It's just doing statistical projections based on averaging over multiple days. You can change the number of days it uses.
Click on the 3 dots in the upper right, then click settings, then advanced preferences.
Speaking very loosely: The smoothing days setting affects the backwards look, i.e., how strongly individual fluctuations affect the trend line. Setting it longer (more days) will tend to make the past part of the line less bumpy. Forecast days affects how much of recent history gets used to guess at the future.
You can play with these settings, and see how it affects the graph. Nothing you change here affects the underlying data you've recorded, so play to your heart's content, in order to understand what the settings do. If you reset back to the original values, the chart will look just the same as it originally did, so you aren't hurting anything by messing with these settings temporarily.
I was fine with the default settings while losing. (I do pretty much understand what they do, so I don't freak out about transitory weirdness.) Now, in maintenance, I prefer using longer settings (30/60 repectively) because it works better for me with uneven eating I do.
Thanks, that's helpful. I also eat unevenly now, with one day well over the others per week, so my weight jumps up the day after for a few days and then it falls back off. I set it to 10 days and it made the trend a lot less jagged and dramatic. Hopefully 10 days isnt too many. It's making me feel better lol. Is there a certain amount that's recommended?0 -
purple4sure05 wrote: »It's just doing statistical projections based on averaging over multiple days. You can change the number of days it uses.
Click on the 3 dots in the upper right, then click settings, then advanced preferences.
Speaking very loosely: The smoothing days setting affects the backwards look, i.e., how strongly individual fluctuations affect the trend line. Setting it longer (more days) will tend to make the past part of the line less bumpy. Forecast days affects how much of recent history gets used to guess at the future.
You can play with these settings, and see how it affects the graph. Nothing you change here affects the underlying data you've recorded, so play to your heart's content, in order to understand what the settings do. If you reset back to the original values, the chart will look just the same as it originally did, so you aren't hurting anything by messing with these settings temporarily.
I was fine with the default settings while losing. (I do pretty much understand what they do, so I don't freak out about transitory weirdness.) Now, in maintenance, I prefer using longer settings (30/60 repectively) because it works better for me with uneven eating I do.
Thanks, that's helpful. I also eat unevenly now, with one day well over the others per week, so my weight jumps up the day after for a few days and then it falls back off. I set it to 10 days and it made the trend a lot less jagged and dramatic. Hopefully 10 days isnt too many. It's making me feel better lol. Is there a certain amount that's recommended?
I can't imagine how there would be a recommended value. It's an interplay between our eating style (very even vs. very uneven, at the extremes), our goals (lose/maintain/gain), and our personal psychology (calm or inclined to freak out), as I see it.
If you can understand statistics enough to actually understand with the app is doing, that's very empowering (but is probably multiple semesters of stats classes, which I assure you are very boring
). Short of that, if you can get the gist of what the values do by playing with them - as you've done - you can try different settings over time, and see what works best for you.
I still sit down sometimes and change mine up, temporarily, just to answer some question I have in my head. 1 -
purple4sure05 wrote: »It's just doing statistical projections based on averaging over multiple days. You can change the number of days it uses.
Click on the 3 dots in the upper right, then click settings, then advanced preferences.
Speaking very loosely: The smoothing days setting affects the backwards look, i.e., how strongly individual fluctuations affect the trend line. Setting it longer (more days) will tend to make the past part of the line less bumpy. Forecast days affects how much of recent history gets used to guess at the future.
You can play with these settings, and see how it affects the graph. Nothing you change here affects the underlying data you've recorded, so play to your heart's content, in order to understand what the settings do. If you reset back to the original values, the chart will look just the same as it originally did, so you aren't hurting anything by messing with these settings temporarily.
I was fine with the default settings while losing. (I do pretty much understand what they do, so I don't freak out about transitory weirdness.) Now, in maintenance, I prefer using longer settings (30/60 repectively) because it works better for me with uneven eating I do.
Thanks, that's helpful. I also eat unevenly now, with one day well over the others per week, so my weight jumps up the day after for a few days and then it falls back off. I set it to 10 days and it made the trend a lot less jagged and dramatic. Hopefully 10 days isnt too many. It's making me feel better lol. Is there a certain amount that's recommended?
Given the way you reacted to the effect in the trend-line caused by a gain of less than half a pound, I would say you should set it to at least 30 days.5 -
lynn_glenmont wrote: »purple4sure05 wrote: »It's just doing statistical projections based on averaging over multiple days. You can change the number of days it uses.
Click on the 3 dots in the upper right, then click settings, then advanced preferences.
Speaking very loosely: The smoothing days setting affects the backwards look, i.e., how strongly individual fluctuations affect the trend line. Setting it longer (more days) will tend to make the past part of the line less bumpy. Forecast days affects how much of recent history gets used to guess at the future.
You can play with these settings, and see how it affects the graph. Nothing you change here affects the underlying data you've recorded, so play to your heart's content, in order to understand what the settings do. If you reset back to the original values, the chart will look just the same as it originally did, so you aren't hurting anything by messing with these settings temporarily.
I was fine with the default settings while losing. (I do pretty much understand what they do, so I don't freak out about transitory weirdness.) Now, in maintenance, I prefer using longer settings (30/60 repectively) because it works better for me with uneven eating I do.
Thanks, that's helpful. I also eat unevenly now, with one day well over the others per week, so my weight jumps up the day after for a few days and then it falls back off. I set it to 10 days and it made the trend a lot less jagged and dramatic. Hopefully 10 days isnt too many. It's making me feel better lol. Is there a certain amount that's recommended?
Given the way you reacted to the effect in the trend-line caused by a gain of less than half a pound, I would say you should set it to at least 30 days.
I'm posting here because I was confused, not because I was freaking out. I was confused because people encourage others to use these apps to understand the overall picture, and now the overall picture is indicating that I'm going to gain 10 pounds by the end of the year just because my weight was up a couple pounds for a few days, which didnt seem realistic. I'm trying to understand how useful these apps are and how to make them the most accurate. Thanks for the suggestion though, I will probably increase it to a larger number of days.2 -
purple4sure05 wrote: »lynn_glenmont wrote: »purple4sure05 wrote: »It's just doing statistical projections based on averaging over multiple days. You can change the number of days it uses.
Click on the 3 dots in the upper right, then click settings, then advanced preferences.
Speaking very loosely: The smoothing days setting affects the backwards look, i.e., how strongly individual fluctuations affect the trend line. Setting it longer (more days) will tend to make the past part of the line less bumpy. Forecast days affects how much of recent history gets used to guess at the future.
You can play with these settings, and see how it affects the graph. Nothing you change here affects the underlying data you've recorded, so play to your heart's content, in order to understand what the settings do. If you reset back to the original values, the chart will look just the same as it originally did, so you aren't hurting anything by messing with these settings temporarily.
I was fine with the default settings while losing. (I do pretty much understand what they do, so I don't freak out about transitory weirdness.) Now, in maintenance, I prefer using longer settings (30/60 repectively) because it works better for me with uneven eating I do.
Thanks, that's helpful. I also eat unevenly now, with one day well over the others per week, so my weight jumps up the day after for a few days and then it falls back off. I set it to 10 days and it made the trend a lot less jagged and dramatic. Hopefully 10 days isnt too many. It's making me feel better lol. Is there a certain amount that's recommended?
Given the way you reacted to the effect in the trend-line caused by a gain of less than half a pound, I would say you should set it to at least 30 days.
I'm posting here because I was confused, not because I was freaking out. I was confused because people encourage others to use these apps to understand the overall picture, and now the overall picture is indicating that I'm going to gain 10 pounds by the end of the year just because my weight was up a couple pounds for a few days, which didnt seem realistic. I'm trying to understand how useful these apps are and how to make them the most accurate. Thanks for the suggestion though, I will probably increase it to a larger number of days.
It assumes the gain will be the result of you repeating the same calorie surplus daily. You can still turn it around. My trend finally shows a 750 average deficit and a goal date of July. It took a few weeks to smooth out. Stick to your plan and it will move downward.1 -
@New_Heavens_Earth Thanks for this! Very helpful explanation!1
-
Hilogirl2018 wrote: »@New_Heavens_Earth Thanks for this! Very helpful explanation!
You're welcome 😊1 -
purple4sure05 wrote: »lynn_glenmont wrote: »purple4sure05 wrote: »It's just doing statistical projections based on averaging over multiple days. You can change the number of days it uses.
Click on the 3 dots in the upper right, then click settings, then advanced preferences.
Speaking very loosely: The smoothing days setting affects the backwards look, i.e., how strongly individual fluctuations affect the trend line. Setting it longer (more days) will tend to make the past part of the line less bumpy. Forecast days affects how much of recent history gets used to guess at the future.
You can play with these settings, and see how it affects the graph. Nothing you change here affects the underlying data you've recorded, so play to your heart's content, in order to understand what the settings do. If you reset back to the original values, the chart will look just the same as it originally did, so you aren't hurting anything by messing with these settings temporarily.
I was fine with the default settings while losing. (I do pretty much understand what they do, so I don't freak out about transitory weirdness.) Now, in maintenance, I prefer using longer settings (30/60 repectively) because it works better for me with uneven eating I do.
Thanks, that's helpful. I also eat unevenly now, with one day well over the others per week, so my weight jumps up the day after for a few days and then it falls back off. I set it to 10 days and it made the trend a lot less jagged and dramatic. Hopefully 10 days isnt too many. It's making me feel better lol. Is there a certain amount that's recommended?
Given the way you reacted to the effect in the trend-line caused by a gain of less than half a pound, I would say you should set it to at least 30 days.
I'm posting here because I was confused, not because I was freaking out. I was confused because people encourage others to use these apps to understand the overall picture, and now the overall picture is indicating that I'm going to gain 10 pounds by the end of the year just because my weight was up a couple pounds for a few days, which didnt seem realistic. I'm trying to understand how useful these apps are and how to make them the most accurate. Thanks for the suggestion though, I will probably increase it to a larger number of days.
Well, I will take a stab at an answer as I may have been one of the more persistent pushers of weight trend apps on the MFP forums.
The primary reason I push them is because people repeatedly freak out by a single high meal / unexpectedly bad weight ins and go into convulsions. Given the nature of water weight changes this is something that happens all the time and these apps dampen down the up and down swings to a level that is more reflective of reality while still detecting persistent upward and downward trends when one looks BACK at them. (I say more reflective of reality because single events don't determine our long term weight; persistent patterns of events do).
I personally use trendweight.com automatically connected to my fitbit.com account, thus allowing me to transfer my weight information to three apps with a single entry. I enter on fitbit and it transfers to trendweight and MFP.
Trendweight takes into account past weight ins and projects forward for something that on my screen looks like a bit less than a week.
Given that I believe trendweight looks back at the past 10 days, I believe that the forward projection extends about 5 days.
They show this forward projection for the two week, four week, and three month view and don't even show one for the six month and longer views.
This makes sense to me as you're taking a chunk of past data and projecting the behaviour a reasonable amount of time forward.
Just like MFP's 5 week projection based on a single day's information is garbage, extrapolating a one year trend based on 5 to 10 days of data is also garbage.
So this is, I think, an expectations issue because we are trying to perceive usefulness on something that a sanity check ought to tell us makes no sense!
I am by no means a statistician (or even mathematically aware as I believe grade 12 high school kids know more math than I do these days); but, gut instinct tells me that if I am making a decision based on 5 days of past data, anything that is predicting more than half that time forward would be of extremely limited value.
Perhaps there is a better app on Android, one with more sane "defaults" in terms of projections shown than Libra. if that's the case I would love to know more about it so I can suggest it to people instead.
In any case, the trend is supposed to be looked at to show us how our general weight level **has been** changing given your current amount of eating. Over periods of PAST weeks, not days.
Retrospectively these apps show you what happened and what your "true" weight trend was. Moving forward they can only take a guess based on things staying EXACTLY THE SAME.
Obviously a few days of over-eating will show up in an upward spike and if you know you've brought that over-eating under control the projection is not showing (and cannot show) anything useful since it doesn't yet have the data that will show that you've brought the issue under control. Nor can the app know that a hormonal upward weight swing has come to an end and the downward one has started for those so affected. Nor can it know that the water weight spike was because of a surgery and it will go down in a few weeks.... etc.0 -
purple4sure05 wrote: »lynn_glenmont wrote: »purple4sure05 wrote: »It's just doing statistical projections based on averaging over multiple days. You can change the number of days it uses.
Click on the 3 dots in the upper right, then click settings, then advanced preferences.
Speaking very loosely: The smoothing days setting affects the backwards look, i.e., how strongly individual fluctuations affect the trend line. Setting it longer (more days) will tend to make the past part of the line less bumpy. Forecast days affects how much of recent history gets used to guess at the future.
You can play with these settings, and see how it affects the graph. Nothing you change here affects the underlying data you've recorded, so play to your heart's content, in order to understand what the settings do. If you reset back to the original values, the chart will look just the same as it originally did, so you aren't hurting anything by messing with these settings temporarily.
I was fine with the default settings while losing. (I do pretty much understand what they do, so I don't freak out about transitory weirdness.) Now, in maintenance, I prefer using longer settings (30/60 repectively) because it works better for me with uneven eating I do.
Thanks, that's helpful. I also eat unevenly now, with one day well over the others per week, so my weight jumps up the day after for a few days and then it falls back off. I set it to 10 days and it made the trend a lot less jagged and dramatic. Hopefully 10 days isnt too many. It's making me feel better lol. Is there a certain amount that's recommended?
Given the way you reacted to the effect in the trend-line caused by a gain of less than half a pound, I would say you should set it to at least 30 days.
I'm posting here because I was confused, not because I was freaking out. I was confused because people encourage others to use these apps to understand the overall picture, and now the overall picture is indicating that I'm going to gain 10 pounds by the end of the year just because my weight was up a couple pounds for a few days, which didnt seem realistic. I'm trying to understand how useful these apps are and how to make them the most accurate. Thanks for the suggestion though, I will probably increase it to a larger number of days.
Yeah, honestly I don't see the point to having it set to such a short time window for calculating the trend, because the argument you hear over and over for these apps is that it helps you see the trend over a longer period of time, even if you're weighing every day. But if you (or the default) set it to such a short period of time that minor variations (much smaller than most people would consider normal daily fluctuations) have such a huge effect on the projected trend, it's not accomplishing what everybody claims is so great about a weight trend app.
Best of luck.1 -
New_Heavens_Earth wrote: »purple4sure05 wrote: »lynn_glenmont wrote: »purple4sure05 wrote: »It's just doing statistical projections based on averaging over multiple days. You can change the number of days it uses.
Click on the 3 dots in the upper right, then click settings, then advanced preferences.
Speaking very loosely: The smoothing days setting affects the backwards look, i.e., how strongly individual fluctuations affect the trend line. Setting it longer (more days) will tend to make the past part of the line less bumpy. Forecast days affects how much of recent history gets used to guess at the future.
You can play with these settings, and see how it affects the graph. Nothing you change here affects the underlying data you've recorded, so play to your heart's content, in order to understand what the settings do. If you reset back to the original values, the chart will look just the same as it originally did, so you aren't hurting anything by messing with these settings temporarily.
I was fine with the default settings while losing. (I do pretty much understand what they do, so I don't freak out about transitory weirdness.) Now, in maintenance, I prefer using longer settings (30/60 repectively) because it works better for me with uneven eating I do.
Thanks, that's helpful. I also eat unevenly now, with one day well over the others per week, so my weight jumps up the day after for a few days and then it falls back off. I set it to 10 days and it made the trend a lot less jagged and dramatic. Hopefully 10 days isnt too many. It's making me feel better lol. Is there a certain amount that's recommended?
Given the way you reacted to the effect in the trend-line caused by a gain of less than half a pound, I would say you should set it to at least 30 days.
I'm posting here because I was confused, not because I was freaking out. I was confused because people encourage others to use these apps to understand the overall picture, and now the overall picture is indicating that I'm going to gain 10 pounds by the end of the year just because my weight was up a couple pounds for a few days, which didnt seem realistic. I'm trying to understand how useful these apps are and how to make them the most accurate. Thanks for the suggestion though, I will probably increase it to a larger number of days.
It assumes the gain will be the result of you repeating the same calorie surplus daily. You can still turn it around. My trend finally shows a 750 average deficit and a goal date of July. It took a few weeks to smooth out. Stick to your plan and it will move downward.
That's what the MFP "complete your diary button" does. Trend apps don't work that way. Different ones probably use different models, but they're all taking data points over a period of time and plotting a line/curve to best approximate the relationship between those data points and points in time, or a moving trailing average of those points, or a weighted average, or weighted moving trailing average ... They're not taking into account your calorie surplus at all, except to the extent that your calorie surplus is reflected in your weight (data points).1 -
It's just doing statistical projections based on averaging over multiple days. You can change the number of days it uses.
Click on the 3 dots in the upper right, then click settings, then advanced preferences.
Speaking very loosely: The smoothing days setting affects the backwards look, i.e., how strongly individual fluctuations affect the trend line. Setting it longer (more days) will tend to make the past part of the line less bumpy. Forecast days affects how much of recent history gets used to guess at the future.
You can play with these settings, and see how it affects the graph. Nothing you change here affects the underlying data you've recorded, so play to your heart's content, in order to understand what the settings do. If you reset back to the original values, the chart will look just the same as it originally did, so you aren't hurting anything by messing with these settings temporarily.
I was fine with the default settings while losing. (I do pretty much understand what they do, so I don't freak out about transitory weirdness.) Now, in maintenance, I prefer using longer settings (30/60 repectively) because it works better for me with uneven eating I do.
Didn't realize we could manipulate this, just changed mine-thanks!0 -
purple4sure05 wrote: »Is there a certain amount that's recommended?
I can't imagine how there would be a recommended value.
This is correct; it's really a matter of personal preference and should be based on your individual goals, time frames, etc.
Personally, I'm eating at a fairly aggressive caloric deficit so I use the last 7 days' data; when I log my weight Wednesday, the previous Wednesday drops from the equation used to calculate the trend line. You could easily extrapolate this to 2 weeks, 30 days, etc. to meet your preference. For example, when I switch to bulking my weight gain will (ideally) be slow, so I'll likely use the last 14-28 days' data.
If you're not familiar with statistics or just generally curious try a few variations and see what you like! I'm also a stats nerd, so feel free to PM me with any questions.2 -
Libra is designed to lag a few days and gives more credit to older weight so it doesn’t project too rapid of a loss . I wouldn’t worry too much about the future line because I don’t think it’s intended to be too precise. You can see it’s a Direct extension of the last few days trend. Just keep following your plan!1
This discussion has been closed.
Categories
- All Categories
- 1.4M Health, Wellness and Goals
- 393.6K Introduce Yourself
- 43.8K Getting Started
- 260.3K Health and Weight Loss
- 175.9K Food and Nutrition
- 47.5K Recipes
- 232.5K Fitness and Exercise
- 431 Sleep, Mindfulness and Overall Wellness
- 6.5K Goal: Maintaining Weight
- 8.6K Goal: Gaining Weight and Body Building
- 153K Motivation and Support
- 8K Challenges
- 1.3K Debate Club
- 96.3K Chit-Chat
- 2.5K Fun and Games
- 3.8K MyFitnessPal Information
- 24 News and Announcements
- 1.1K Feature Suggestions and Ideas
- 2.6K MyFitnessPal Tech Support Questions