Nominal, Ordinal, Interval, Schmordinal
19 November 2014
Don't teach significance testing – Guest post
2 February 2015

Risk is an important topic

In order to make informed decisions about screening and medical interventions, people need to have a good understanding of risk and probability. The communication and understanding of risk was a very popular topic at the ICOTS 9 Conference. I have written previously about risk, but want in this post I wish to introduce our new video about risk and screening, and talk more about the communication of risk.

Human elements of risk

When we teach about screening for disease, we need to be aware that many of the things we are screening for, particularly forms of cancer, can have emotional connections in our students. We may not know that one of our students has a family member who is dying of cancer. It is important that we teach about screening for cancer, but it can be a rather depressing or even trigger an emotional upset. Teachers may know their students’ circumstances and deal compassionately with this, and always speak sensitively of disease and incidence.

Ear Pox

Camilla's ear pox was not treated in time.

Camilla’s ear pox was not treated in time.

When we made our video we were aware that people in all different circumstances would view it. In order to be able to use our usual light-hearted approach in our video, we decided to avoid talking specifically about real-life diseases and their consequences. Rather we invented a disease called ear-pox, which has very convenient values for prevalence, sensitivity and specificity. The outcome of untreated ear-pox is that the ear falls off. (Which is easy to show in an animation.) The outcome of a false positive was that a person’s ear was tattooed unnecessarily. We hope that this provides a semi-realistic example, that is not upsetting to people.


An icon array can make the proportions easier to visualise.

An icon array can make the proportions easier to visualise.

In the video we use two different ways of representing the information – an icon diagram and a natural frequency table. Both formats have their strengths in helping people to understand the implications of the figures. Because we wish the layperson to have an understanding and “feel” for the implications of the figures, it is very important to find ways to represent the situation that resonate. There is a body of research into the evolutionary history of quantitative information that suggests that people are better able to understand frequencies – things we can count, than proportions. The picture of six green dots and four red dots is more intuitive than being told that the proportion of success is 60%. Icon diagrams use dots of different colours and outlines to indicate the different states in question. This is becoming a popular way to express these concepts. You can see an animation here of the risk associated with eating bacon.  Being able to see what one person in 100 looks like, is powerful in terms of defusing anxiety regarding incidence. Here is a link to an icon array used to illustrate the effects of breast cancer screening.
The table of natural frequencies is another representation that aids in understanding. This turns probabilities into numbers of people who are in each of the four categories – Correctly identified as affected, correctly identified as not affected, and false positives and false negatives. Where incidence is low, which it often is, the number of false positives outweighs the number of correct positives by a large margin, and this is shown well on the table. Tables can be used more easily than diagrams to calculate the probability, for example, that a positive result is false.

Teach about false positives

When we teach about probability and risk, it is important to make clear the negative impacts of a false positive diagnosis. These can have lasting effects of people’s health and well-being. In my work I spend quite a bit of time on a plane, and when I am not reading Amish romances I get to talk to all sorts of people. One very interesting conversation was with a genetic counsellor. As my son has a severe disability as a result of a pair of autosomal recessive genes, my husband and I had once visited such a counsellor, and I knew of their purpose. In this single-serving plane relationship, we got to talking about people’s perceptions of risk with regard to genetics, which I found fascinating. The genetic counsellor said she had talked to people who were horrified at a one in one thousand risk of some adverse outcome. In contrast other clients were relieved that the probability of an outcome (like the one for my son) was only one in four. The perceived impact of the probability is of course tempered by the severity of the result, and the worldview of the people concerned. It is also affected by their perception of independence in probabilistic outcomes. Unfortunately there are still people who think that with having had one child with a one-in-four outcome, the chances are increased that the next three children will be fine.

It is important

Teaching people about risk, independence and probability is a holy work. We can help people to make informed choice about their own health and that of their children. The Harding Center for Risk Literacy and Sir David Spiegelhalter and his colleagues are doing great work. I would love to hear of other websites that we can link to – please add them in the comments. I hope that our new video can likewise contribute.


  1. majmurr says:

    Why not use a real animal disease (like bovine TB) rather than a phony human disease? Real data could be obtained and the transition to human diseases is not hard to make. Murray Jorgensen

  2. Geoffrey Brent says:

    I’m doing an online genetics course at the moment, and the lecturer often uses one particular real-life disease as an example. As it happens that disease has hit my family very hard, I’ve lost three close relatives to it, which makes it hard not to get distracted when that one comes up.
    So I think the approach you take above is a good idea (likewise Murray’s alternate suggestion).
    – Geoffrey B

    • Dr Nic says:

      Hi Geoffrey. Thank you for sharing that with us. Maybe you could pass the idea on to the instructor. Coming from Christchurch, I am very sensitive to discussions about earthquakes, which I’m sure most people would find harmless. It is good for instructors and examiners to think about the impact of their choice of example.
      – Nic

  3. When making a video, you were probably aware that people in all different circumstances would view it.

Leave a Reply

Your email address will not be published. Required fields are marked *