Tamsin E. Lee

This catchphrase is dangerous. As a lover of digging into datasets to uncover the stories, this ubiquitous catchphrase irks me. Data cannot capture the full story, and expectations that it can polarizes people.

Reality is beautifully complex with infinite connections and nuances that, to replicate, needs both an infinite amount of data and an infinite number of analysis monkeys. Stating that decisions should be based on data gives data analysis a bad name.

Mathematicians do not rule the world

People in power need to lead, inspire, negotiate, and be diplomatic. It is not necessary to know about Bayesian probability to succeed in their role. In fact, mathematicians have a bad rep at schmoozing. (How can you identify an outgoing mathematician? They look at your shoes instead of theirs.)

It is not just people in power. Most people do not have the kind of training to intuitively understand mathematical models or statistical analysis. And yes, it would be great if everyone did. But that’s not going to happen any time soon. And whether we all could is a conversation for another blog entry.

Most people I’ve met recognize the importance of maths models and data analysis (albeit my sample is very biased!), and they’re willing to listen. But if the results suggests something that is against their known reality, it discredits maths models and data analysis.

This disconnect in a vital part of the process, and it not acknowledged for the necessity and enlightening process that it is. Actionable insight is a team effort across different experts. Give me a dataset and I can happily play with it for hours. But my real joy comes from creating something with people who know about the reality –  I learn from them and in turn we both learn from the data.

Let’s create something beautiful together!

Imagine water flowing on the ground. There’s not rocks nor anything to guide the water flow. I see my role as creating a form from this. I talk to people to learn, “Oh this isn’t possible? Okay, I’ll put a rock here so the water can’t go there.”, “Oh this is an important dynamic? Okay, I’ll dig a trench here to make sure these components are connected.”. Not every sculpting effort will be perfectly placed initially. But as a joint project, we can create an elegant water feature that takes us from A to B. And in my fortunate experience, this process engages and captivates everyone involved.

Towards a future with “Data informed decisions”

Let’s update the catchphrase! Any model, any analysis, is not a blueprint. It is only information. Information that cannot be obtained any other way. So it is vitally important. Moreover, the learnings gathered during this creative collaboration mean that critical thinking is slowly but surely permeating into decision making systems, which benefits us all!

Below is a post written by Shweta Sharma, a Masters student from May 2020 to July 2021 whom I supervised at the Swiss Tropical and Public Health Institute.

Take it away Shweta!

It’s spring of 2020, and I am starting my masters  thesis under the Disease Modelling Research Unit at Swiss TPH. Everyone is talking about COVID-19, but I will be talking about malaria! With Tamsin, I will investigate drug resistant malaria in India.

It seemed like a perfect and easy topic for me. I am from India and being with a medical background, it would be is easy for me to understand everything.

But… there is much more. There is geography, mathematical modelling, and coding.

“It’s easier to resist at the beginning than at the end.”  – Leonardo da Vinci

Malaria parasites develop mutations that lead to resistance to antimalarials. India treats malaria with Artemisinin plus Sulfadoxine-Pyremethamine (SP), which requires the parasite (Plasmodium falciparum) to develop at least five mutations before the drug is no longer effective. The World-Wide Antimalarial Resistance Network (WWARN) has gathered 652,83 samples from 180 publications from 62 countries to map where SP drug resistant mutations have been detected. I used their data for India for my thesis.

Not all states in India have data on SP drug resistance. So, we consider factors that are associated with antimalarial drug resistance, and make predictions for the states without data (using regression analysis). See the map below, which goes from partially coloured to completely coloured.

From countless to countable

There are countless factors associated for drug resistance (predictor variables). To make an initial selection of ‘predictor variables’, I read. A lot. (I love doing literature reviews. You read and learn, read and learn). I learnt that socioeconomic factors are a big driver of drug resistance in India. For example in rural areas, Unlicensed Medical Practitioners (UMP) are preferred for seeking treatment and they do not follow the national malaria treatment guidelines. I settled on 44 predictor variables that covered socioeconomic factors, epidemiological factors and health infrastructure. I collected data for these variables, at the state level, from different official sources in India.

To choose which of these 44 predictor variables are the most relevant, I used combinations of two or three predictor variables in a regression model to predict the presence of drug resistant malaria for states where we know the level of drug resistance. And as a control model, I also used random values as predictor variables (the null model). The combination of predictor variables that gives predictions that most closely match the actual data are the winners.

States with low socioeconomic status are more likely to have drug resistant malaria

The three most accurate combinations were, in order:

1. Average population per government hospital and per capita income

2. Average population per government hospital and gross enrolment ratio (Gross enrolment ratio is the ratio of enrolment in higher education to the population in the specific age group. In the data I used the age group between 18 to 23 years for both genders.)

3. Average population per government hospital and the percentage of the population that lives in a rural area.

Average population per government hospital is very important as rural and urban areas have disparity in terms of the quality of health care services. Rural health centres are overburdened resulting in compromised quality of care. Combining this variable with a socioeconomic factor, makes the predictions more accurate than using the average population per government hospital as the only predictor.

From data to predictions

The predictions don’t match the data exactly, clearly seen with Andhra Pradesh state (yellow on left map). However, the predictions are overall very close to the data, and we can fill in information where we don’t have data.

What I learnt

Thesis – I completed my master’s thesis and learnt that socioeconomic factors are predictors of the presence of drug resistant malaria.

Geography – Now I can spot all Indian states on a map 😛

Maths – I can understand big equations better.

Coding – I can make pretty maps in R.

Tamsin, I will be grateful always for what I have learnt from you…you are an amazing person and a great mentor !

Imagine there’s a new disease. Quickly, stock up on loo roll! But whilst we panic, the powers that be throw money at the problem (whilst, presumably, also panicking). Then knowledge grows and miracles happen.

Miracles happen because we invest in learning

The ‘learning curve‘ is the line that shows how knowledge increases with experience. For this blog post, I’m going to replace ‘experience’ with money, meaning that as more money is invested, more knowledge is gained. (In maths terms, “For disease management, assume experience is proportional to the amount of money invested”.)

The Knowledge here (capital K) ranges from all scales of understanding that is needed to eliminate a disease: from the small biology stuff, to the development and production of tests, treatments, and vaccines, to mathematically modelling individuals’ behaviour, to delivery of equipment and medicine, and to provision of care throughout the at-risk population. The goal is to obtain aaaallll the Knowledge (maximum Knowledge). (Maths note, Knowledge is a dimensionless parameter, where the relevant aspects are its relationship to investment, and the comparisons of the maximum Knowledge between the different diseases.)

For this toy example, I’m looking at a snapshot in time so I exclude anything that changes with time, such as new variants. Even with this simplification, managing diseases is already complex. But in 2022, I doubt this is news to anyone.

The learning curve above has three stages

• 1. Scoping: A lot of money is needed, with relatively little returns. There is a lot we don’t yet know.

• 2. Blooming: Things are rolling, and investments return a lot of Knowledge. We’re feeling bloomin’ good!

• 3. Trudging: There’s diminishing returns on investments. This last slog can make it difficult to justify further investment, especially as the number of diseased people would have decreased rapidly thanks to the blooming stage. However, without trudging through this final stage, diseases can resurge, and thus cost more in the long run.

The new disease occupies the scoping stage of the learning curve. Whereas an older disease is past this stage, and occupies a region that may encompass the last, trudging stage. So at a snapshot in time, the ‘Age’ (capital A) of a disease is an indicator of how close we are to having the disease under control (from herein, I’ll refer to this ideal point as eliminating the disease).

Where to invest?

Suppose you rule a world with these two diseases, one old and one new. The currency of this world is a yip (your portrait is on each yip note, perhaps pulling a different face for different denomination). At this snapshot in time, how should you plan to allocate your yips between the two diseases?

Let’s find out with some maths and pretty plots!

I’m investigating the importance of the Age of the disease, so I will run simulations with parameter choices that highlight this effect (see the inner plot of the figure below). Specifically,

• A: Both diseases have the same learning curve, but the section that is currently occupied by them is different. In reality, diseases vary, so one disease may have a longer scoping stage, or a shorter trudging stage, or any other variation. (In maths terms, changing the value for parameter a of the Sigmoidal curve.)
  
  
• B: At this snapshot, the new disease occupies the scoping stage and the first half of the blooming stage. The old disease occupies the latter half of the blooming stage and the trudging stage.

The plot below presents the answer, which is in terms of the

• bankroll: a percentage of the total amount of money needed to eliminate both diseases (to gain maximum Knowledge).

• allocation percentage: the percentage of the bankroll allocated to each disease.

What if the new disease can affect many more people?

I ran a scenario where the new disease is more disasterous than the old disease. Specifically,

• A*: the new disease requires three times more Knowledge than the old disease, and costs twice as much as the old disease to eliminate. These values reflect that Knowledge includes provision of care throughout the at-risk population, so a larger population at-risk means more Knowledge is required. At the same time, a larger population at-risk doesn’t necessarily mean the lab research is more costly, so the new disease is only two times more expensive to learn about.

The role of toy examples

In reality, disease management is much more complex, so fortunately we would not completely ignore a disease. This complexity can lead to benefits not discussed here, such as an overlap of returns. For example, strengthening the health system to manage an existing disease will benefit management of new diseases.

Nonetheless, this toy model demonstrates that when managing multiple diseases, where they are in their individual learning curves is relevant. That is, how readily do we expect Knowledge about each disease to be gained, and where are we in that trajectory?

Also, toy models are fun! Play with me! My R code is available on GitHub. You can add more diseases, change the learning curves for each disease, change the section of the learning curve that each disease occupies, and add noise to the learning curves. Packages used are dplyr, ggplot2 and reshape2. (Methods: The optimum percentage of the bankroll allocated to each disease was calculated using the genetic algorithm, a global optimiser. The genetic algorithm was run for each percentage of the available bankroll.)

I’m excited to announce the publication in Malaria Journal, the leading journal in malaria research,

Mechanistic within-host models of the asexual Plasmodium falciparum infection: a review and analytical assessment.

Flavia Camponovo, Tamsin E. Lee, Jonathan Russell, Lydia Burgert, Jaline Gerardin, Melissa A. Penny

The lead author is my good friend Flavia Camponovo, who was a PhD student in the same research group as me (she is now at the Communicable disease dynamics, Harvard). She had a keen interest in developing a model of the within-host dynamics of a person infected with malaria. Like any good scientist, she started collecting information about the current within-host models. However, she’s a very thorough scientist, so instead of writing a paragraph summarising the current state of research, she went much further and started reproducing a lot of the current models, coding them all in Matlab. Granted, some codes were provided, but as anyone who has tried to reproduce just one paper knows, this is a huge task! So in the way that research can take unexpected directions, reproducing and comparing current models became a paper in itself. This deep dive has been gracefully executed, thereby providing a lot more than a typical review does.

Needing to be needed

As mentioned in a previous post, for nearly two months I had to work from a campsite in a locked down New Zealand. This unique experience taught me that I’m not the type of person who produces her best work when isolated on a beautiful beach. I’ve learnt that I need a windowless cave to work hard! When I did manage to resist the lure of the sea and instead sit at my laptop, I looked at my work for 10 minutes before compulsively updating COVID stats and repatriation information. Fortunately, it was at a time when Flavia’s paper was coming together, and she sent me a copy of her paper. What motivates me is working with other people, and knowing that others depend on my input. So I engrossed myself in the paper, grateful for a tangible task. Still, I was frustrated that my lack of concentration meant I was reading each sentence two or three times. I had little faith in my comments because I was returning to work after three months off. I no longer felt like a scientist but more like a hippy traveller.

A picture is worth a thousand words

I was befuddled when she later text me a photo of a beautiful diagram she was making, which was inspired by one of my comments, also shown in her photo. As I squinted to read my comment, I didn’t recognise it. It sounded so sciency! (I was requesting information about how each model included feedback between the immune response and infected red blood cells.) I didn’t sound like a hippy whose daily routine included emptying a camper toilet and dancing in the sea. What’s more, I didn’t envision that my comment could be addressed with a diagram. Flavia took my comment, agreed with it, and addressed it using her knowledge and skillset.

What a delicious ping-pong between scientists 🙂

Papers. Oh they take some time. I am pleased to announce a new paper (well newish. Published September 2020 Tropical Conservation Science). This paper started in 2018, but it’s always been a side project so I’ve struggled to make the time to make it a focus.

I am happy to finally have this off my to do list, but I’m more happy to share the work with you here. This is from work I’ve been doing with one of my favourite collaborators, Dr. David Roberts at The University of Kent.

Illegal trade of wildlife is very difficult to monitor. Fortunately this is Dave’s expertise! One approach is to find patterns in legal online trade, so we suggest an analysis that gets the most out of this data.

Maximising what we know

Our dataset is from a study of online wildlife trade which investigated the trade of Convention on International Trade in Endangered Species animals (including live animals, parts and derivatives), over 280 open online marketplaces, across 16 countries, during a six week period in 2014 (Hastie & McCrea-Steele, 2014).

Our study was considerably more extensive than usual data on legal wildlife trade, which is currently not collected with a global mindset. Instead there is a focus on the extensively traded items from countries which trade large amounts, for example, China trades ivory products and Germany trades turtles and tortoises. However, by applying some statistical analysis, we can shine a light on the less dominating behaviours.

Looking at 31 different types of wildlife products, traded by 16 different countries, we cluster the products according to their trade patterns (using hierarchical clustering in R), so products which are traded similarly are grouped together. The 31 products were grouped into eight categories, five of which grouped items which were predominately traded by one country. For example, cluster 1 contained ivory, rhino and pangolins, which are all largely traded by China. During the time period of the dataset, 1662 ivory items were traded by China, but only 164 rhino items, and a measly 3 pangolins. With this distribution, it is understanding that reports tend to focus on the ivory items. However, by grouping them together, one can make assumptions about the illegal trade of all of these items for the price of one!

Other products do not have a strong association with a particular country (from our list of 16 countries). These products were exotic birds, primates, otters, antelopes, red panda, conches, cats, crocodiles and alligators, foxes, bears and whales, see the figure below. Therefore monitoring the illegal trade of these items would require surveillance in many countries.

Voice and Accountability

Returning to the five clusters which contain wildlife items that were traded by a dominating country. These countries were China, Germany, the Netherlands, the UK, and Poland. (The trade of items from the other three clusters were not dominated by a particular country.) Ordinating these clusters we looked for a relationship between the items traded and Worldwide Governance Indicators, which are six measures for each country that include factors such as control of corruption and government effectiveness.

These indicators come with controversy, of course! I personally find them nice measures for understanding the world. They are defined here (Kaufmann et al., 2010). We found a correlation between the items traded and the ‘Voice and Accountability’ score for each country. The  ‘Voice and Accountability’ captures perceptions of the extent to which a country’s citizens are able to participate in selecting their government, as well as freedom of expression, freedom of association, and a free media. So using this correlation (not causation!) we can make trade predictions for countries not included in our data. For example, comparing the Voice and Accountability score for the United States (a country not included in our dataset), with the average Voice and Accountability score for the clusters, we infer that the United States traded elephant items (not ivory) and owl items during 2014 (items from cluster 6).

What pets do you have?

Sharing the results of this paper with my international colleagues was fun, as my German and Dutch colleagues learnt that Germany and the Netherlands hosts Europe’s largest reptile trade shows, Hamm Terraristik (as well as its associated online trading platform) and Terraria-Houten. As my Dutch colleague said, “I don’t know anyone with a pet turtle…”. I guess she’s not asking the right questions at parties.

My last blog post, I announced that I was awarded the Marie Curie Individual Fellowship from the Horizon 2020 fund from the European Commission. And I promised to give details about the actual project. Et voila!

Antimicrobial resistance

You’ve probably heard of antibiotic resistance? It’s when the bacteria that’s caused an infection in someone is resistant to the treatment used to kill it off. It’s scary stuff. It’s not just bacteria that develops resistance to treatment. Viruses and parasites and fungi do too. The umbrella term is antimicrobial resistance (AMR). And AMR is a big deal! When a treatment stops working, morbidity and mortality from the disease goes up. New drugs need to be developed, which takes a lot of time and money. And even if new drugs were as easy to develop as a new tea flavour, to be effective to the resistant strain, the new drugs are often more toxic, leading to harsher side effects. And often the new drugs are more expensive, which can make them inaccessible to many patients. The consequence of AMR are so far reaching, that addressing AMR will help 6 out of 17 of the Sustainable Development Goals set by WHO.

What makes a microbe develop resistance? Nature. Consider a sick person, and the offending microbe is multiplying like crazy within this person. Mutations of the microbe occur all the time, and VERY occasionally some of these mutations will lead to a resistant strain. When the settings favour this resistant strain, such as when the sick person gets treatment which kills off all other strains, the resistant strain has an evolutionary advantage. The sick person remains sick. And the infectious disease continues to spread. The resistant strain of it. Dun dun dunnnn…

Drug resistance in low to middle income countries.

To prevent drug resistance, treatment needs to be effective and the full prescribed course of treatment needs to be completed. However, this can be difficult to monitor, especially in low to middle income countries where health care access may be limited. For example, someone may feel better after a few days of treatment, and save their medicine for a family member who has the same symptoms. Or perhaps treatment is cheaper from a family friend with boxes of cheap (presumably counterfeit) drugs. These factors are known to promote drug resistance occurring and spreading.

Monitoring drug resistance spreading

Consider malaria, a parasitic infection. The latest treatment used to combat malaria is artemisinin. Like all antimalarials that came before artemisinin, drug resistance to this treatment starts in South East Asia. When drug resistance malaria infections occur in Africa, where malaria prevalence is very high, the tragedy rapidly escalates. Consequently, drug resistance is monitored globally. The map below is from many molecular marker studies, beautifully collated together by WorldWide AntiMalarial Network (WWARN). These studies take samples from infected people and identify whether the malaria parasite within them has mutations which are associated with resistance to treatment.

My project – the science bit

To demonstrate my project, consider a toy example in Square Land over 20 years. Within Square Land you have sampled infected people, at different locations and times, and identified whether they carry a resistant strain or not (like the WWARN data above). In addition, suppose you know the location of five treatment access points, such as health care centres or hospitals. The first and last year look like this.

The black circles are the treatment access points. A red square means that at that location and time, there is at least one infected person carrying a resistant strain (y_i = 1). A blue square means that at that location and time, there were no sampled people carrying a resistant strain (y_i = 0). In total there are N data points (y₁, y₂, … , y_i, … , y_N), collected over the 20 years. The ‘true’ number of people carrying a resistant strain will be greater, this is only the collected data. In other words, each data point y_i has a probability p_i of being a 0 or 1, where p_i depends on the true density of people infected with a resistant infection at that particular location and year, u_i, and a probability of being sampled. The sampling probability depends on demographics about the person, such as their age.

The true number of people with a resistant infection at a particular location and year, u_i, is unknown. However, we assume some basic knowledge about the underlying processes which determine this truth. That is, we assume that the true number of people with a resistant infection at a particular location and year, u_i, depends on (1) the number of people with resistant infections in the neighbouring regions, and (2) the number of people with resistant infections at the particular location during the previous year. Both of these components depend on the prevalence of the disease in general. Processes (1) and (2) are standard assumptions regarding species spread – dispersal and growth. I added a third process which is specific to drug resistant infections: (3) new introductions of the resistant strain – which is an underlying spatial component that states there is more chance of having a resistant strain when close to a ‘hot spot’, where each hot spot has a different magnitude of resistant infections that it contributes into the population. For simplicity, in this example, I previously stated that all these hot spots are treatment access points, however one or more of them could be a transport hub such as train station or airport. Also, for simplicity, I’ve assumed that we know the location of the hot spots (however this isn’t a requirement of the method).

Using Monte Carlo Markov Chain with uninformative priors (translation: using powerful statistical tools combined with computing power and assuming no previous knowledge – which could introduce biases), we can identify which hot spot is contributing more resistance to the population and target strategy accordingly. In this example, we identified that the top left hot spot contributes the most resistant infections into the population. Therefore, if this hot spot was a health care centre, it would be worth investigating the quality of the drugs administered here and/or the adherence of the patients. If this hot spot was a transport hub, it would inform us that drug resistance is entering the population from outside.

Some of you may be disappointed that I haven’t presented the beautiful equations that describe this whole process. If you want all the juicy details, feel free to reach out or peruse a recent presentation I gave at the Research Center for Statistics at the University of Geneva, available here.

My project – the me bit

I’ve had a varied research background (see previous blog post: From electricity to malaria). This meant that from my PhD, I’m familiar with partial differential equations (used here to model the mechanisms of the true number of people with a resistant infection at a particular location and year). From working with ecologists, I’m familiar with Bayesian hierarchical models being used to model animal movement, which formally accounts for uncertainty in the data . And from working with epidemiologists, I’m familiar with drug resistant malaria. Put it all together…. and what do you get!? I believe the first time that this type of model is applied to understand the spatio-temporal mechanics of AMR in a population.

My goodness! It’s been over a year since my last post?! I’d question what happened, but I think we all understand that 2020 has done funny things with our perception of time.

The hiatus is particularly long on my part because I took three months leave to travel New Zealand, and then I got stuck for an extra two months on a campsite during lockdown – not knowing when and how I would get back to Europe. Although even without a proper toilet and shower, there are worse places to have a lockdown than a friendly campsite that has direct access to a beach…

Upon my return it’s been a series of unfortunate events. Each one on it’s own devastating, but not unbelievable. However, the sheer rate of these events was unbelievable. This series taught me that in order to move forward, I have to meet my reality where my reality is.

So in the spirit of acceptance, I am finally ready to admit to myself, and the internet, some amazing news which I’d been scared to share because I couldn’t believe it…

* * I was awarded the very competitive Marie Curie Individual Fellowship from the Horizon 2020 fund from the European Commission. HUUUZAAAARRRRRR!!!!!! * *

The funding supports talented (and lucky) individuals to work on a research project of their choice for two years. Like all research funding, the proposed project must be state-of-the-art yet achievable. The Marie Curie fellowship stands out because there is a strong focus on supporting the development of the fellow so that by the end of the two years I’ll be captivating you with my dazzling-brain and eloquent explanations… These expected achievements seem unbelievable but hopefully it will become a reality that I’ll meet 😉

This success is due to the incredible support from Prof. Melissa Penny, and a jar of fancy peanut butter with agave syrup and cinnamon.

What’s the actual project? Next blog post coming in the next few weeks…
(Provided no more unfortunate events!)

I’ve been vegan for 18 years. When I went vegan there were rudimentary cheese alternatives. This didn’t both me as I would eat cheese made from potatoes whilst singing, “Potato cheese, potato cheese, cheese made from… potatoes!” Others tasted it and screwed up their face whilst I chomped down on a cheese and Marmite sandwich. Or a cheese and pickle sandwich. Oh salty joy! Then the novelty wore off. Vegan cheese is expensive. And I returned to my first loves, peanut butter and hummus. Although not together. I tried this once and it wasn’t great.

Many years on,  and veganism has risen. There are some vegan cheese options which make potato cheese look like potato. But I haven’t craved cheese, so I’ve largely ignored the options. Until two weeks ago when I was offered a free sample of a faux ‘Tom’ cheese made from fermented cashews. I hated it. It was fermenty and strong and, well, cheesy. Is this imitation so good, and 18 years without cheese means that now I hate cheese? Or alternatively, is this vegan cheese simply bad?

H0: Vegan cheese tastes good (but I don’t like cheese).
H1: Vegan cheese does not taste good.

I ran two experiments with my colleagues as the test subjects.

Experiment one

I left out two cheeses at work, without their labelling.
Cheese 1: The same vegan Tom cheese that I was given a sample of.
Cheese 2: A normal Tom cheese.
I asked colleagues to state which cheese they preferred.

The results were simple. Only 14% (2/14) preferred the vegan cheese. (This wasn’t rigorous because people could see the responses, so social influence has an effect.)

But wait! Perhaps this vegan cheese was particularly bad. So back to the shop I go…

Experiment two

Again, at work, I left out cheeses without their labelling. This time I left out four cheeses.
Cheese 1. Vegan cheese A
Cheese 2: Vegan cheese B
Cheese 3: Dairy, lactose free cheese
Cheese 4: Normal Tom cheese

Both vegan cheese A and B were also made from fermented cashew nuts, but they were not the same as the original vegan cheese that started this all. Now I asked for a ranking system. The results are shown below, where the cheeses are presented in order of popularity.

Fig 1. The ranking of the four cheeses, ordered by most popular to least.

The popularity order was determined by comparing the mean score. And out of interest, three pairwise comparisons confirm that the top three cheeses ranked similarly:
1. Normal cheese with Vegan A cheese (Cliff’s delta -0.19),
2. Normal cheese with Lactose free cheese (Cliff’s delta -0.27),
3. Vegan A cheese with Lactose free cheese (Cliff’s delta -0.05).

Pair wise comparisons with Vegan cheese B were not required as it is a clear loser here.  From the Cliff’s delta score, Vegan cheese A and Lactose free cheese B are almost equally popular.

So it seems that vegan cheese can be enjoyed nearly as much as dairy cheese, especially if it’s lactose free dairy cheese. But vegan cheese varies a lot. Out of the three vegan cheeses my colleagues tasted, only one could compete for the taste buds of non-vegans. So did I like the popular vegan cheese? Well no. In fact, I could tolerate Vegan cheese B most easily, perhaps because it had less flavour.

In conclusion, I reject the null hypothesis. Even though some vegan cheeses can win over cheese lovers, overall, the fermented cashew lacks the cheesy goodness that my colleagues crave.

Why deprive myself of cheesy goodness? To answer that I have to describe my consumer mindset, which is mostly driven by my crippling indecisiveness.

Three dimensional shopping

I’m a dithering and stressed shopper. Few moments in my life have filled me with self-doubt more than staring at 30 options for tweezers in Boots. I realised that I am a novice to the decision factors involved in tweezer shopping.

If I was Neo, the film would have ended with me sitting on the floor, Googling the difference between all the options.

Instead of tweezers, let’s consider the simpler task of clothes shopping. Because when choosing whether or not to buy a new clothing item, my decision process used to be boiled down to (i) desire (ii) affordability, see Fig 2(a).

Fig 2. Every circle represents a new item of clothing that gets my attention. A surrounding square represents a purchase. Plot (a) represents how I used to shop, weighing up desire and affordability only. Plot (b) represents how I shop now.  The third dimension, ethics, is represented by a linear scale from unethical (dark red) to ethical (dark green).

Suppose every circle here represents a new item of clothing that, for a brief moment, gets my attention. And every surrounding square represents that I brought the item.
When shopping, my indecisive brain is overloaded so I used to buy cheap things unquestionably in order to avoid making decisions about whether or not I truly wanted them. And very rarely, I’d buy an expensive item that I really, really wanted.
With only affordability and desire as my explicit decisions factors, I’d walk out of Primark with bags of shopping. I knew this shopping process was creating a demand for dangerous working conditions, and I’d feel a bit of guilt, but guilt doesn’t stop me wanting something!

A few years back I started adding the ethics of the item as a third dimension.
As consumers, we all have our ethical boundaries. Making mine explicit as a decision factor restrains my dithering a bit.

Suppose all the items before can be coloured from green to red representing ethical to not (respectively). Now I shop more like Fig 2(b).

With the separation of ‘desire’ and ‘ethics’ I don’t buy cheap, unethical, things unless I superdouperreallytruly want them. The consequence is that I own fewer pointless, pretty things. Which is perhaps a downside, because put a pretty print on anything, and I want it. Yes. Anything.

On the upside, I’ve indulged in a few more spectacular items that come with twinges of smug joy. And on a confused side, my loved, unethical, purchases come with twinges of guilt each time they’re used. Which in many cases, is daily. Fortunately they’re so pretty that they distract me from my inner-conflict.

A sparkly bag is my bacon sandwich. The joy is far too great for any other thoughts.

The non-linearity of ethics

Despite my representation of ethics on a linear scale of unethical to ethical, it’s impossible to quantify the ethics of a lifestyle. I once spent a month in Bolivia primarily eating Oreos, which although vegan, contains palm oil. It’s unlikely that my Oreo month was the right thing for the world.

And I will continue to make unethical choices daily, so I can’t judge, and I don’t wish to be judged. Ethics and personal choices are complex and impossible to get right. We all have our own decision factors to balance. My choices relate to me, the many privileges that I have, and my easily-overwhelmed indecisiveness. I can’t get it all right, but veganism seems like an easy step towards something that is, at least, in the right direction. So although I miss out on cheese, and whatever other taste sensations warrant pity for vegans, I choose my simpler life over and over. And as it turns out, I don’t even like cheese! So I guess I’m not a martyred vegan. I’m a happy vegan!