Regression Analysis

Whydahdad71

Hello, I was wondering if anyone here has ever used statistical analysis to predict weight loss? I am thinking of using a regression analysis model to see if it could be a tool that helps accurately predict weight loss...just thinking aloud.

sgt1372

Probably over-thinking the problem.

You already know that wt (as the dependent variable) varies directly w/cal intake and exercise or cal expenditure (which are the two major independent variables that affect wt).

What more do you need to know?

Do you really need to do a regression analysis to *predict" what your wt will be if you vary these independent variables?

If so, go for it but don't be disappointed of your results don't exactly fit the model.

PAV8888

Kevin Hall has already done a model (as I am sure have others).

And it is not a straight 3500 per lb.

And you will have a hard time separating fat vs lean mass lost and their contributions.

And there are errors everywhere when it comes to estimating calories in and calories out.

And a retrospective calculation will be valid for you, specifically, but as a sample of one will be of limited use to the rest of us.

And very few people succeed in maintaining a large weight loss for a long period of time.

And in all this doom and gloom we also do know that starting from a position of maintenance and lowering intake and increasing expenditure in the ~8-12% range for each of them works wonders. (Increased expenditure can be achieved by increasing activity, increasing exercise, or both. It is not a necessary condition to achieve weight loss or even weight management, though, generally, it would be considered conducive. Your biggest bang for the buck will be control of your caloric intake)

The rest will vary. And absolute accuracy, or more than "gross" predictability is not necessary for success. What you do with your caloric balance MOST of the time will determine your end results!!!

Along the lines of your original question check out the body weight planner (advanced mode) and Kevin Hall's associated paper in terms of predicting weight loss per application of deficit.

At the same time check out other papers that essentially say that long term maintenance will require ongoing work and commitment.

Which is where I come in to say: don't make it too hard on yourself and try to have fun along the way. It's definitely worth it if you're starting out from a position where your weight has become a health concern!

Duck_Puddle

Will your regression model account for the myriad of behavior changes that need to happen to facilitate most weight loss? Not everyone needs to make wholesale changes to achieve loss, but many do.

Aside from all the things @PAV8888 mentioned, human behavior is what drives the results. For the most part, we know what we need to do to lose. We may not have the minutia figured out, we may not have figured out the specifics of a way we will enjoy eating, we may not have figured out the precise level and type of activity that is right for us-but we all know (deep down in our diet-pill buying hearts) we need to eat less and move more than we have been if we want to lose weight.

Those who lose are the ones who have found a way to modify their behavior to do those things in a lasting and consistent way.

That’s the predictor. What makes people change their behavior in a lasting way?

AnnPT77

I'm assuming (?) that weight loss (pounds or percent or rate or something) would be your dependent variable?

What kinds of things do you think would be candidate independent variables?

Why do you think regression analysis is the right tool? How do you think the regression model you are contemplating would complement the statistical underpinnings we already rely on: The stats behind TDEE/NEAT/BMR estimates, METS-based exercise estimates, tracker-based or HRM-based personalized calorie expenditure estimates, weight trending apps?

Just curious.

Whydahdad71

I am not sure that it would be. I am just thinking about it as "another approach" that could be fun to implement.

happysquatter

I say go ahead and do it. If you like data, it’s nice to play around with it. I’m always fiddling around with my own training, nutrition and body fat data.

Keep in mind, it’ll only apply to you and your body will keep changing so the models predictive value won’t be that strong

If nothing, everyone needs more Excel in their life. I say this seriously.