Visualizing Difference

How to (properly) tell if a number has changed


Visualizing Difference is a one-day course for NHS information analysts that covers several statistical techniques designed to show if a difference between two numbers really is a difference. Each technique is taught using real-life healthcare examples. And the examples are linked together by a single, unified narrative that unfolds as the course progresses. We use Microsoft Excel for the hands-on coursework.


Two overlapping bell curves
This is a chart showing two overlapping Normal distributions. The bell-shaped curve on the left represents a random sample of days from a year when Emergency Department (ED) attendances averaged 330 per day. The curve on the left represents a random sample of days from the following year when ED attendances increased to 346.5 per day. The calculations summarize some of the steps we need to take in order to ascertain if that difference of 16.5 attendances per day was a statistically significant increase.

Session 1: Differences between Proportions

The first session starts with an explanation and discussion of the Null Hypothesis. We then explore three different ways of testing for - and visualizing - differences between proportions: P-values, confidence intervals and the chi-squared test.

Session 2: Differences between Means

In the second session, we look at testing for the difference between means when the underlying data distributions are Normally distributed. We explore how to test for Normality using Q-Q charts, and then go through two ways of testing for statistically significant difference betwen means: the Large Sample Normal test and the t-test.

Session 3: Difference between means (when the underlying data is not Normally distributed)

In the third session we explore how to transform non-Normal data to see if we can make it Normal. Additionally, we investigate the extent to which a P-value obtained from log-transformed data differs from a P-value obtained using the Large Sample Normal test. We also provide a brief overview of bootstrapping as a technique that can overcome the problem of non-Normality.

Session 4: the Mann-Whitney U test and the Chi-squared Test>

In the final session of the course we embrace non-parametric tests. By ranking the data, we can test for difference by using the Mann-Whitney U test. And we can also bring the Chi-squared test into play when we need to compare more than two proportions.


The course teaches the visualizations with a series of themed emergency care examples, so that the relevance of the techniques can be more easily grasped. The examples have been selected so that they are follow-able by people unfamiliar with the acute hospital environment. Every teaching example and exercise uses NHS data that has been used in real situations to shed light on real problems. This is not a course about statistical graphics for their own sake; it is about using visualization to make sense of real issues.



Visualizing Statistics can be delivered as either an on-site, in-person, face-to-face workshop OR as a virtual course via Microsoft Teams. In either case, the cost is £1,250+VAT, and up to 12 participants can be accommodated. Email info@kurtosis.co.uk to start making arrangements.