The FastDay Forum

[carorees]

Calling all fellow nerds, particularly those with abilities on the statistical analysis front!

So, now we have a lot of data on the progress tracker and a good number of members who have been following the diet for a while. Is there anyway we can tease out what differentiates the quick from the slow in terms of weight loss? We can see a fairly normal curve developing around the 0.5 kg/week point but there are definitely two subgroups of people at either end of the spectrum.

There are so many factors that influence weight loss but if we focus on modifiable factors, we might be able to offer constructive advice. Such factors might include: numbers of fast days/week; fast day calories (total cals or relative to TDEE); fast day meal organisation; fast day meal components; feed day calories (relative to TDEE); feed day diet (e.g., low/high carb/fat/protein); exercise frequency/intensity. Any others?

How could we collect data to try to answer those questions? I was wondering whether we could use the people who contribute to the monthly weigh in, divide them into quartiles by weight loss (or waist reduction) and send them a questionnaire? We could try to see if there were any features common to the top quartile not found in the bottom quartile and vice versa.

Thoughts my nerdy friends?

I can't believe you just posted this as when I saw your latest stats I was just wondering the very same thing.
I think the key might be the non fast days.
I notice a lot of posts about this.
If you could solve the problem of how much is too much, and how little is too little, then we'd all be losing weight every week, and we could probably make a fortune into the bargain!

This would make an excellent dissertation subject as there is a cohort of subjects and a decent hypothesis to test!

I think the questionnaire could be very useful, but I have been on 5:2 for 6 weeks now, have only lost 2lbs, but in that time have changed just about all the variables you mention; feed days food style/calorie intake, exercise levels, etc etc. So I couldn't even say if I have been within TDEE over that period, or followed the low-carb diet I'm theoretically on, or even how much exercise I've done, as its varied so much from week to week. I am pondering how to work out the impact of the different changes i make, but its so hard when weight changes have so many variables (water retention etc etc) to pin down whether a change has had a positive impact

Well it could use stuff from MFP?

I think both me and Minumonline both said we track what we eat on MFP??

It's something I am toying in my own head with, writing a spreadsheet which takes calories in, and calories out, and matches a predicted weight loss against actual, then a variance analysis could be done.

MFP tracks calories, Net Calories, and Carbs protein and fats.

I for one would to happy to share this info with "friends" on MFP.

The thing that I really need to get a handle on is actual calories burned.
if you put in that you have a sedantary lifestyle and then try and count your exercise it seems impossible, and normal heart rate fitness monitors arn't that good so it seems.

Fasting today, just had 3 x 30 calorie Cauliflower Pizza's and I feel FAT!!!

How?

You're right. MFP is the answer - it not only shows the weekly calorie intake, it also shows the fat:carb:protein ratio, split over days, and what we have entered as exercise. So pretty much all the variables we need. If we put our fast day food in as meals, we could even use it to analyse the differences between having a single fast day meal, or splitting into 2/3 small meals.

I think that covers all the variables we've discussed?

MFP is really good but not sure how many people use it. I suspect the amount of calories eaten on feed days is key to how quickly weight is lost. Getting this info may show whats going on.

It may also be an idea to filter out people like me that are no longer trying to lose weight, from the overall data.

[carorees]

My reasoning for suggesting dividing the population into quartiles and then comparing the top quartile with the bottom one was so that we would be comparing big losses against little losses and so perhaps the exact calorie count or exact weight change would not matter. If we had enough data we could divide the population even finer and look at just the top and bottom 15%!

Although the scale readings may be inaccurate as long as they are equally inaccurate each time the change in weight should be evident! I would suggest looking at the losses over a reasonable length of time to avoid problems with day to day variation.

As far as the TDEE calculation goes in some ways we could just look at the calories against a standard TDEE calculator or perhaps a BMR calculator so we could look at exercise separately? We could just report our findings as the calories relative to the 'standard' TDEE calculation.

I think people using MFP or fat secret calorie counters would be the best way to get a lot of the info we need. Again it might not have to be too accurate in the first instance.

So what about a pilot study? We recruit people who are already using MFP/fat secret, or are willing to start using it, who undertake to record details for a month and not change their way of 5:2ing during that time. At the end of the month we split them into quartiles by weight loss and gather as much info as possible from those in the top and bottom quartiles to see if there are any clear factors. We would need perhaps to try to match the groups as far as possible for age, gender, BMI, and time on diet as we know they will be confounders.

I think we should try to get people who have already had at least a month on 5:2 to avoid the impact of those first weeks which seem to be very variable and may have something to do with water retention and speed of adapting to the lifestyle (a subject for an other project!). Our findings will perhaps help us to work out how to drill down further into what is going on!

Do we have any statisticians who can run a multiple logistic regression on the data? If not perhaps we can learn something with a simple statistical calculation? We can't assume a normal curve so it will have to be a non-parametric test? I am rubbish at statistics so I need advice here!

Once we have finalized the study design we can start recruiting volunteers!

Multiple linear regression would be a start, and you can include log (data) to help with curvy things. Excel does multiple regression and I'll have a look what else is readily available.

[Tracieknits]

I think it sounds intriguing!! I'm going to have to bow out though, because I have a very identifiable factor which makes weight loss slow - I take thyroid slowing medication for my hyperthyroidism. I actually think my dose might be too high presently and I've gone for bloodwork to have this checked. I've been feeling awfully cold in my fingers and toes! A too-slow thyroid is an enemy to weight loss.

[carorees]

Oh thanks Phil...I relied on the hospital statistician to do my stats for my PhD so I'm fairly hopeless!

[carorees]

You may indeed be right but we need to test what makes the difference for sedentary middle aged women not compare them with you, hence the aim to match the populations if possible and to perform a regression analysis so we can find the independent variables. I am a middle aged and sedentary woman and yet my average weekly weight loss has been around 1.5 lb+ which is twice what some other middle aged sedentary women are achieving. What am I doing different from similar people who are not losing so fast?

Also the progress tracker suggests 4:3 does not actually give much boost to weight loss but we don't know whether that is because of differences between the population doing 5:2 and 4:3 or a real difference.

On the face of it the simple mathematical formula you give does not work out in practice. Two people apparently similar in terms of TDEE are not getting the same weight loss it seems. The idea of my research proposal is to try to find out what differences between the two are leading to this disparity.

As you say there are a lot of variables, many of which cannot be changed (e.g. age, gender, height, starting weight) but if we can tease out which are the modifiable variables, then people can, well, modify them!

Do sedentary middle aged women of similar BMI show a wide range of loss rates ?

[carorees]

My impression is yes, but our survey/study will tell us. The progress tracker does not show the range just the median (unless I've missed it) so we don't know.

I am in.

I would love it if we were able to answer some more questions.

Personally I do not believe a man and a woman, with the same height, weight and exercise would loose weight any differently if eating the same foods.

I would like to prove me right! ( I doesn't matter if we don't as I know I am right anyway!!!!)