Weight gain and lockdown: empiricism vs. physical modelling

Summary

Life in lockdown has disrupted our daily lives, creating the perfect setup for putting on pounds. In a poll of 1,000 people, nearly half of the women and almost one-quarter of the men reported that they had gained weight “due to COVID-19 restrictions”. I presume all 1,000 have good, quantitative observational time-series of their weight change.

Weight gain

Sleep cycles have been ruined, anxiety is rising and many people are falling into an unhealthy lifestyle. I certainly score three-out-of-three on that list.

In recent days I had been hoping to process (curve-fit) the Zoe (citizen science) symptomatics time-series for different regions across the UK, and to write a blog page about the results. However, I have been unable to gain access to the Zoe database. In practice, despite the scheme’s encouraging claims about sharing data or providing access to the ‘deidentified’ data, one is endlessly passed from one database manager to another, or presented with yet another e-form to fill in. The net result is that my 14-day-old application has so far drawn a complete blank.

I had planned to combine Zoe fits with hospital deaths data and so be able to back-calculate the current COVID-19 infection rate, Ro, and the state of regional progression towards ‘herd immunity’. These two key factors are needed for planning the lifting of lockdown. Instead I am reduced to analysing my weight change!

The graph below shows my observations, plus a smooth (empirical) fit, of my weight in 2020. I began by slowly shedding some unwanted bulk (doctor’s orders) in January/February/early March. COVID-19 caused me to begin a stringent (99%) social isolation regime (commencing 6th March). My weight almost immediately started to fall off more rapidly. Official lockdown saw a slight slowing of my weight-loss. (I place the start of lockdown as 23rd March when the UK government banned all “non-essential” travel, shut down almost all businesses, venues, facilities, amenities and places of worship). Unfortunately, in recent days I have started to regain weight.

Time-weight graph. Vertical lines mark dates of personal and official lockdowns. Target weight 82 kg. When, if ever, will I reach my target weight?

In truth, I doubt if any of the wiggles and bumps in the red curve (a lowess – LOcally WEighted Scatterplot Smoother) actually have anything to do with lockdown, despite some interesting looking associations. Instead the vacillations illustrate the difficulties of interpreting a purely empirical fit (purely data-driven fit) in comparison with the far better situation of using a model with some underlying physical or physiological basis.

Empiricism vs. physical modelling

To return to the main theme of this series of blog postings – namely applying Hubbert’s maths to COVID-19 time series data – I am always much happier with a model than an empirical fit. This is especially the case when prediction (i.e. extrapolation) is involved. Hubbert based his predictions of peak oil and ultimate resource on a model underpinned by a linearized cumulative logistic equation.

It’s quite sobering to recognise how little is revealed by the weight-change analysis. Without a model it’s almost impossible to judge how my weight will change in the future, or when my target weight might be reached. Mechanistic, or deterministic, modelling is much better than mere empirical fitting. Analytical and deterministic models are to be preferred over empirical models as they are much more general and broader in application. They will often have linear, exponential, or logarithmic ‘backbones’ which can be selected using classic statistical techniques – or as I do with Hubbert through an analysis of variance.

Two articles

In closing. I have been reading two contrasting reports about COVID-19. I found both to be enthralling.

A Greek tragedy

One by the health editor of the Guardian worries, like me, that a second wave, or possibly worse, a long extended tail is in the offing – if the Government continues to treat the virus as a political crisis rather than a medical emergency. Sarah Boseley has won a number of awards for her work on HIV/Aids in Africa. In her article she writes about the views of Robert West, a professor of health psychology at University College London’s Institute of Epidemiology and Health, who sits on the advisory group on behavioural science for Sage. Neither could be exactly classified as a Cummings fan. She writes
“what we now have to recognise is that central government probably cannot be trusted to provide leadership” and furthermore the whole Cummings odyssey “is playing out like a Greek tragedy with the protagonists trapped by their own proclivities to self-destruct – in his case by his hubris and inability to say sorry.”

The end of the affair

The other is by Alistair Haimes who adopts the diametrically opposite point of view. He writes in the Critic “It is the end of the affair. We are no longer at epidemic levels of covid-19 prevalence in the UK … Parents, unions and nervy adults fret about the risk, but there is little need.”

Both articles make fascinating reading.

Glimmer of hope XXII / Roy Thompson by blogadmin is licensed under a Creative Commons Attribution CC BY 3.0