Coronavirus, lifestyle diseases and the Shadow Mean

In this post, I introduce fat-tailed distributions and the concept of the Shadow Mean, with implications to how seriously multiplicative events should be taken in the society. [Addendum: If you want a technical treatment of the proper Shadow Mean approach instead of my caricature, see this]

I keep getting struck by how often we see well-meaning educated people comparing phenomena such as terrorism and epidemics to the “as much or more” dangerous lifestyle diseases. I even saw one of the smartest health psychologists I know commit this error in their professorial inauguration speech. Note, that I’m not against preventing non-communicable diseases; in fact, that’s what my dissertation is about. But we need to be vigilant on how risks work.

Here’s a chart from the aforementioned presentation, where you can clearly see that, all else equal, we should be diverting almost all our prevention resources to the biggest killers, which are lifestyle diseases:

Rik causes of death

The problem is, that all else is not equal. Why?

It has to do with a concept called “Shadow Mean” (capitalised for ominosity), which relates to “fat tailed” distributions. I’ll explain more later.

But let us first consider some properties of the Coronavirus pandemic, and how they differ from the common flu – and, by extension, to lifestyle diseases. To do so, I’ll give the floor to Luca Dellanna (Twitter, website), who kindly contributed his thoughts to this blog:


Luca Dellanna: Six unintuitive properties of the current pandemic

1/6: Asymmetry (part I)

“The cost of paranoia is bounded. The sooner we get paranoid, quicker we can get a handle on things, sooner we can confidently go back to business as usual the cost of “letting it happen” is unbounded. Here is the tradeoff in the US: Restrict international travel now and maintain our ability to move freely domestically or keep the flows coming and inevitably have to restrict movement both internationally and domestically. The choice is clear.” – Joe Norman (link)

There is enough evidence that the pandemic is inevitable. The only question is how big and how fast we want it.

The costs of preventing the pandemic are mostly linear. Closing down schools today for one month costs roughly as much as closing them for one month in April. Closing down 3 schools costs roughly half as closing down 6 (assuming the same size).

Instead, the costs of letting the pandemic grow are nonlinear.

Letting the pandemic run today might mean 100 more people infected tomorrow. Letting the pandemic run next week might mean 1000 more people infected the following day.

And it gets worse (see the next point).

2/6: Nonlinearities

“In the US, we have 2.3 million people in prison. I cannot imagine a way to stop #coronavirus from spreading like wildfire among that population. How will federal, state, & local authorities handle this?” – Jon Stokes (link)

Another example of the non-linear consequences of the pandemic.

A pandemic that “knocks-off” (i.e. prevents from working, for any reason) 0.1% of the workforce is bad but not that bad.

A pandemic that “knocks-off” (i.e. prevents from working, for any reason) 0.1% of the workforce in a clustered way is much worse: it means that some companies lose a large percentage of their workforce for a few days or weeks and must close the operations (whereas others are directly unaffected).

A pandemic that “knocks-off” (i.e. prevents from working, for any reason) 0.2% of the workforce is ten times worse than a 0.1% pandemic – for there are less workers to covers those who are sick, for one company closing creates problems downstream the supply chain, and so on.

The worst case is so bad that it makes sense planning for it even if it has low chances to happen (which is itself a strong assumption on too uncertain variables).

3/6: Impact

“The difference between the flu and the coronavirus is that between a tide and a tsunami. The same amount of water, but the impact is different because the tsunami arrives all at once.” – Roberto Burioni (link)

As I explained on Twitter, the problem is not (only) the current mortality, but the mortality we can get if our healthcare system gets overwhelmed. People won’t receive the care they need, even for conditions unrelated to the coronavirus.

“If a juggler can juggle 4 balls letting them drop 1% of time,  then he can also juggle 10 balls letting them drop 1% of time.” – this is how most people estimate mortality. As if healthcare was a fully elastic system.

4/6: Asymmetry (part II)

“Asymmetry. Convex decision. So long as there is no risk of harm from masks & disinfectants, the decision is wise, in spite of the absence of evidence– Nassim Nicholas Taleb (link)

Face masks do not offer full protection, but they do offer some protection. As long as you remove them carefully and they don’t make you sweat (so that you’re tempted to touch your face), they’re better than nothing.

Their cost is minimal and bounded, their benefit is large and unbounded (at least for you: they might save your life).

Of course, there is the argument that face masks are finite and they should be allocated where they’re the most needed. It’s a valid argument. But let’s focus on the asymmetry of the cost-benefit, because it applies to another method as well: washing hands and disinfecting.

Their cost is extremely low. I’m baffled that so few people are doing it first thing while arriving home.

Don’t be penny-wise but pound-foolish with your time.

5/6: Testing

“True epidemic in Iran and South Korea, community spread in Italy, confirmed transmission from Iran to multiple countries, the US basically isn’t testing anybody… and as far as I can tell it’s gauche even to mention [the virus] in public in the United States.” – @toad_spotted (link)

If a country doesn’t like to talk about a problem, it will have to talk about that problem.

Problems grow the size they need for you to acknowledge them.
The virus is already here, it’s just not evenly detected. – Balajis Srinivasan (link)

6/6: Infection

“I just realized that when people say ‘yeah but you won’t die’ they mean ‘yeah you’ll become a carrier and make everyone you see sick but not die’.” – Paul McKellar (link)

There are many replies to “the coronavirus is not that mortal”.

  • “15% mortality in older people (80+ years old) almost means a Russian Roulette if they get infected”.
  • One’s chances of dying depend on the number of infected people he meets in his day-to-day (because the more he meets, the more the chances he gets the virus).
  • We don’t know! There are many reasons that prevent us from pinpointing the mortality of the virus in a way that is predictive of the future. We should assume the worst scenarios until we can rule them out. (Why? Because asymmetry and nonlinearities; the content of points #1 and #4 above.)

Luca

[Luca’s newsletter is pretty much the only one I’ve ever found positively thought-provoking; if you want to hear more of his ideas, subscribe here]


 

Horizontally challenged tails

What does this have to do with lifestyle diseases? Well, while the incidence of the common flu is quite unlikely to quadruple from one year to the next, it is much, much less likely, that the incidence of e.g. cardiovascular disease would do the same.

Let’s look at an example. In the left plot below, you see what a mortality rate from a fat tailed distribution would look like. There are two years, when you have an extreme case – something psychologists are used to just eliminating from the data. Note, that outliers are different from extremes; an outlier may be a badly measured observation, whereas an extreme lies within the conceivable boundaries of the phenomenon.

fat and thin tails
Figure by me; code available here

The left plot could signify a viral epidemic. Say we are living year 26; the mean observed annual mortality would be around 900, and you probably aren’t too worried; things are almost exclusively very calm. But, given the fat-tailed distribution, extreme values are possible and upon surviving year 27, the mean would be almost 6000. Before it’s seen, this is known as the Shadow Mean; there are yet unobserved cases we can infer from the mechanics that produce the fat-tailed distribution, but which are not (yet) observed empirically.

Contrast the situation with that on the right plot, which could signify deaths from accidents in a country like Finland. In 900 years, we still have not observed one with over 2500 deaths (nb. this is just simulated data from a thin-tailed distribution). The mean is about 1000 and if we omit the maximum observation, it remains practically identical.

lawnmowers
Figure by Stefan Gasic; see his work here!

N-th order matters

Time and second-order effects – that is, things that happen as an indirect consequence of an event – are of great importance when something extreme happens. Let us run a small scenario. Finland has 5½ million people. Let us consider that 25% would get infected (with a maximum of, say, 50%), and 5% (max. 20%) would require care in a hospital. This would already mean, that we would suddenly have 70 000 (max 550 000) extra patients in the healthcare system, which has been “streamlined” for years. Very different scenario than having the same number of extra patients over the course of a year or a decade – one, which lays fertile ground to second-order effects. These include the impact on people, who wouldn’t have big problems under normal situations, due to having hospital care capacity readily available.

Finally: This is not fearmongering or a call for hysteria. Cold-headed rational decision making calls for taking precautions here. If you stock up so that you can self-quarantine yourself for 14 days in the case of getting ill, and do it gradually by buying little extra every time you go to the store anyway, you are making a good decision. Here’s one more figure by Luca, illustrating the point:

Image
Figure by Luca Dellanna; source

Relevant resources and references:

 

Complexity considerations for intervention (process) evaluation

For some years, I’ve been partly involved in the Let’s Move It intervention project, which targeted dysfunctional physical activity and sedentary behaviour patterns of older adolescents, by affecting their school environment as well as social and psychological factors.

I held a talk at the closing seminar; it was live streamed and is available here (on stage starting from about 1:57:00 in the recording). But if you were there, or are otherwise interested in the slides I promised, they are now here.

For a demonstration of non-stationary processes (which I didn’t talk about but which are mentioned in these slides), check out this video and an experimental mini-MOOC I made. Another blog post touching on some of the issues is found here.

 

blogiin kuva

Misleading simplifications and where to find them (Slides & Mini-MOOC 11min)

The gist: to avoid getting fooled by them, we need to name our simplifying assumptions when modeling social scientific data. I’m experimenting with this visual approach to delivering information to those who think modeling is boring; feedback and improvement suggestions very welcome! [Similar presentation with between-individual longitudinal physical activity networks, presented at the Finnish Health Psychology conference: here]

I’m not as smooth as those talking heads on the interweb, so you may want just the slides. Download by clicking on the image below or watch at SlideShare.

SLIDE DECK:

misleading assumptions 1st slide

Mini-MOOC:

 

Note: Jan Vanhove thinks we shouldn’t  become paranoid with model assumptions; check his related blog post here!

The secret life of (complex dynamical) habits

It was recently brought to my attention that there exist such things as time and context, the flow of which affects human affairs considerably. Then there was this Twitter conversation about what habits actually are. In this post, I try to make sense of how to view health behavioural habits from the perspective of dynamical systems / complexity theory. I mostly draw from this article.

Habits are integral to human behaviour, and arguably necessary to account for in intervention research 1–3. Gardner 1 proposes a definition of habit as not a behaviour but “a process by which a stimulus generates an impulse to act as a result of a learned stimulus-response association”. Processes being seldom stable for all eternity, a complex dynamical systems perspective would propose some consequences of this definition.

What does it mean, when a process—such as habit—is stable? One way of conceiving this is considering the period of stability as a particular state a system can be in, while being subject to change. Barrett 4 proposes four features of dynamic system stability, in which a system’s states depend on the interactions among its components, as well as the system’s interactions with its environment.

corpus clock
Corpus Clock of Cambridge, where I’m writing this. The clock behaves chaotically so that it’s accurate every five minutes. A time-eating locust on top reminds us that neither habits, nor other human endeavours, escape this passage. Photo: Jim Linwood

First of all, stability always has a time frame, and stabilities at different time frames (such as stability over a month and a year) are interdependent. We ought to consider, how these time scales interact. For example, some factors which determine one’s motivation to go to the gym, such as mood, fluctuate on the scale from minutes to hours. Others may fluctuate on the daily level, and can be influenced by how much one slept the previous night or how stressful one’s workday was, whereas others fluctuate weekly. Then again, some—which increasingly resemble dispositions or personality factors—may be quite stable across decades. When inspecting a health behaviour, we ought to be looking at minimum the process which takes place on a time scale one level faster, and one lever slower than the one we are purportedly interested in 4. For example, how do daily levels of physical activity relate to weekly ones, and how do montly fluctuations affect the weekly fluctuations? Health psychologists could also classify each determinant of a health behaviour, based on the time scale it is thought to operate on. For example, if autonomous forms of motivation 5 seem to predict physical activity quite well cross-sectionally, we could attempt to measure it for a hundred days and investigate what the relevant time-scales of fluctuations are, in relation to those of the target behaviour. Such an exercise could also be helpful for deciding on the sampling frequency of experience sampling studies.

Second, processes in systems such as people have their characteristic attractor landscapes, and these landscapes can possibly be spelled out, along with the criteria associated with them. By attractors I mean here behaviours a person is drawn to, and an attractor landscape is the conglomerate of these behaviours. The cue-structure of the behaviours can be quite elaborate. For example, a person may smoke only, when they have drank alcohol (1) in a loud environment (2), among a relatively large group (3) of relatively unfamiliar people (4), one or two of whom are smokers (5); a situation where it is easier to have a private conversation if one joins another to go out for a cigarette. This highlights how the process of this person’s smoking habit can be very stable (mapping to the traditional conception of “habitual”), while also possibly being highly infrequent.

Note: Each of the aforementioned conditions for this person to smoke are insufficient by themselves, although all are needed to trigger smoking in this context. As a whole, they are sufficient to cause the person to smoke, but not always necessarily needed, because the person may smoke in some more-or-less limited other conditions, too. These conditions can also be called INUS (referring to Insufficient but Necessary criteria of an Unnecessary but Sufficient context for the behaviour) 6. Let that sink in a bit. As a corollary, if a criterion really is necessary, it may be an attractive target for intervention.

Third, the path through which change happens matters, a lot. Even when all determinants of behaviour are at a same value, the outcome may be very different depending on previous values of the outcome. This phenomenon is known as hysterisis, and it has been observed in various fields from physics (e.g. the form of a magnetic field depends on its past) to psychology (e.g. once a person becomes depressed due to excess stress, the stress level must be much lower to switch back to the normal state, than was needed for the shift to depression; 7). As a health behaviour example, just imagine how much easier it is to switch from a consistent training regime to doing no exercise at all, compared to doing it the other way around. Another way to think about is to consider that systems are “influenced by the residual stability of an antecedent regime” 4. As a consequence of stability being “just” a particular type of a path-dependent dynamic process 4,8, we need to consider the history leading up to the period where a habit is active. This forces investigators to consider attractor patterns and sensitivity to initial conditions: When did this stable (or attractor) state come about? If interactions in a system create the state of the system, which bio-psycho-social interactions are contributing to the stable state in question?

Fourth, learning processes such as those happening due to interventions usually affect a cluster of variables’ stabilities, not just one of them. To change habits, we naturally need to consider which changeable processes should be targeted, but it is probably impossible to manipulate these processes in isolation. This has been dubbed the “fat finger problem” (Borsboom 2018, personal communication); trying to change a specific variable, like attempting to press a specific key on the keyboard with gloves on, almost invariably ends up affecting neighbouring variables. Our target is dynamic and interconnected, often calling for coevolution of the intervention and the intervened.

It is obvious that people can relapse to their old habitual (attractor) behaviour after an intervention, and likely that extinction, unlearning and overwriting of cue-response patterns can help in breaking habits, whatever the definition. But the complex dynamics perspective puts a special emphasis on understanding the time scale and history of the intervenable processes, as well as highlighting the difficulty of changing one process while holding others constant, as the classical experimental setup would propose.

I would be curious of hearing thoughts about these clearly unfinished ideas.

  1. Gardner, B. A review and analysis of the use of ‘habit’ in understanding, predicting and influencing health-related behaviour. Health Psychol. Rev. 9, 277–295 (2015).
  2. Wood, W. Habit in Personality and Social Psychology. Personal. Soc. Psychol. Rev. 21, 389–403 (2017).
  3. Wood, W. & Rünger, D. Psychology of Habit. Annu. Rev. Psychol. 67, 289–314 (2016).
  4. Barrett, N. F. A dynamic systems view of habits. Front. Hum. Neurosci. 8, (2014).
  5. Ryan, R. M. & Deci, E. L. Self-determination theory: Basic psychological needs in motivation, development, and wellness. (Guilford Publications, 2017).
  6. Mackie, J. L. Causes and Conditions. Am. Philos. Q. 2, 245–264 (1965).
  7. Cramer, A. O. J. et al. Major Depression as a Complex Dynamic System. PLoS ONE 11, (2016).
  8. Roe, R. A. Test validity from a temporal perspective: Incorporating time in validation research. Eur. J. Work Organ. Psychol. 23, 754–768 (2014).

 

Evaluating intervention program theories – as theories

How do we figure out, whether our ideas worked out? To me, it seems that in psychology we seldom rigorously think about this question, despite having been criticised for dubious inferential practices for at least half a century. You can download a pdf  of my talk at the Finnish National Institute for Health and Welfare (THL) here, or see the slide show in the end of this post. Please solve the three problems in the summary slide! 🙂

TLDR: is there a reason, why evaluating intervention program theories shouldn’t follow the process of scientific inference?

summary

Getting Started With Bayes

This post presents a Bayesian roundtable I convened for the EHPS/DHP 2016 health psychology conference. Slides for the three talks are included.

bayes healthpsych cover

So, we kicked off the session with Susan Michie and acknowledged Jamie Brown who was key in making it happen, but could not attend.

start

Robert West was the first to present, you’ll find his slides “Bayesian analysis: a brief introductionhere. This presentation gave a brief introduction to Bayes and how belief updating with Bayes Factors works.

I was the second speaker, building on Robert’s presentation. Here are slides for my talk, where I introduced some practical resources to get started with Bayes. The slides are also embedded below (some slides got corrupted by Slideshare, so the ones in the .ppt link are a bit nicer).

The third and final presentation was by Niall Bolger. In his talk, he gave a great example of how using Bayes in a multilevel model enabled him to incorporate more realistic assumptions and—consequently—evaporate a finding he had considered somewhat solid. His slides, “Bayesian Estimation: Implications for Modeling Intensive Longitudinal Data“, are here.

Let me know if you don’t agree with something (especially in my presentation) or have ideas regarding how to improve the methods in (especially health) psychology research!