Then there's the right part of the chart - it struck me that roughly aiming for somewhere around 1,450 kcals a day is going to be less effective as I lose weight, because the number of kcals my body needs to just move around (or to 'maintain' my weight) will decrease as there's less mass to carry. So, I've used the BMR formula (Base Metabolic Rate) for women, entering my weight that week, my height and my age, into the equation. To work out your 'daily calorie needs' from that, you use the Harris Benedict equation, which takes your BMR and multiplies it by an activity factor. I couldn't decide which category I fit into, so I picked a multiplier in between two categories based on the general level of activity I do; in my case, 1.5. So then I used the chart to take my BMR multiplied by 1.5, for my 'daily kcal needs'. The next column gets a little geeky - remember that everything here is very generic, but the 'general' understood belief about kcals and weightloss states that every 1lb of weight is equivalent to 3,500 kcals... so to lose 1lb a week, I need to create a shortfall of 500 kcals per day from my calorie needs. I've used my chart here to work out the difference between my weekly kcals intake, and my weekly 'kcal needs', and then divide it by 3500. So the right hand column shows what 'science' thinks I should have lost that week.
Analysing that, it doesn't really go along with what happened in real life, but that's because of all the other factors around weightloss - natural daily body weight fluctuations, the effect of my menstrual cycle to weigh-ins (through water retention etc.), perhaps how stress/sleep has affected things that week, and the fact that it's blindingly obvious from just reading my blog that I've been way more active some weeks than others, and of course activity isn't really factored in here. But then it gets interesting when you total it up: My actual weightloss for 22 weeks is remarkably close to what the equations say it should be!!! So maybe it's not all as daft as a generic formulaic approach seems like it should be! :-)
Anyway. I can clearly see that when I've eaten more, I've lost less weight (or indeed gained a small amount) - and it gives me a strong indication that aiming to stay around 1450-ish per day remains a good way forward for me, even though the weightloss will steadily decrease week on week as I go on.
So, as it's probably only Robert still reading at this point - what do you think?? ;-)
 |
| as always, click to enlarge |
Very interesting ! I always think you that food affects your weight and exercise your mind. Always good to have the combo. Having read every diet book known to man and especially recent ones that say a calories isn't calorie is so interesting to see your experiment of one showing actually a calorie IS a calorie. Hmm lots to think about. Thanks
ReplyDeleteHaha, experiments of n=1 can also be known as "BroScience"... ;-) Dude, 3500 kcals totally equals 1lb!!!
DeleteAnd yes, re 'a calorie isn't a calorie', of course the above analysis doesn't include how the kcals breaks down to the macro level - fat, protein, carbs - or which weeks included 'junk meals' and so on. Though I'm amused that last week (which included the Half Marathon) and I felt like I stuffed my face with rubbish for pretty much the whole week, kcals wise I still averaged just over 2000!
DeleteGreat post, slightly embarrassed about the title though... ;-)
ReplyDeleteBUT.... where's the graphs!!! Can't do a geeky post without graphs, tables don't do the data you've collected justice. It would be good to plot average daily weight loss vs average daily calories calories for each week, then them same graph again be against average calorie deficit.
In theory the later graph would roughly go through the 0,0 point but looking at your data it doesn't look to, as your estimated daily requirement looks to be lower than the formulas suggest for your weight. Potentially you could skew the estimate by multiply it by a scale factor so the you get the graph heading through the zero point. This scale factor would the means of calibrating these generic formula to your own body + methods of estimating calorie intake. Once you know this scale factor it should be possible to better predict weight changes based on estimates of food intake.
I had a limited lunchbreak time to pull that table together, lol. No time for faffing with graphs (also I'm no good at them on excel). Yes, according to the calorie needs calculations, I should have lost weight every week, but there have been 3 weeks where I haven't. There's definitely a trend of correlation, but not a direct one.
DeleteIt's difficult to calibrate to my own body, because I have a 29 day hormonal cycle to factor in, it definitely plays a role but not the same effect every month. Additionally, my exercise differs week on week, and even if I did the same number of sessions, I couldn't really compare intensity levels. I AM trying to track those weeks against average number of steps on my pedometer each day, which I also faithfully record - I should have that info/analysis finished within a couple more days. :-)
Calibrating models to fit your own data isn't always straight forward so takes a bit of patience with experimenting with different adjustments till you find a scheme that works consistently - or is right more often than the basic models. The fact that the models suggest that you should have lost weight hints to me that it needs to be calibrated, even without accounting for the amount of exercise and the effects of monthly fluctuations.
DeleteWhat I tend to do with analysis is doing it iteratively, first get the basic data together and just plot some graphs to look at general trends/relationship. Sometimes the data is so noisy that as is there isn't anything more meaningful to draw from it that to accept that there are lots of compounding variables. When you see broad trends you can attempt to model this, or adjust models to fit it better.
Once you have these broad trends accounted for then it can be easier to spot other types of relationships - such as monthly cycles, or effects of exercise. Once you have a better handle of this you can make the original adjustments more sophisticated and then reapply them to the data.
With these "loosely controlled" experiments of one you have to be careful about reading too much into the what the data might suggest, it can help to know a bit about the underlying physiology as to know when something is expected or unexpected. These unexpected findings can sometimes provide deeper insight, but other times it's just some artefact of data collection, analysis or some other compounding variable that you've not accounted for.
I am a bit of uber geek but times when I've followed the maths or analysis and come up with some unexpected results it forces one to go back double, triple check and then sometimes you realise that it's all correct and your understanding of a problem can transform completely - like suddenly in 3D or hearing for the first time. These moments I guess what all scientists live for, those epiphanies that provide a deeper understanding of the world around us.