Fig. 3

Multiple Correspondence Analysis plots of 1125 cases, 349 probable cases and 4750 non-cases, showing relationships between symptom variables and disease status in individuals presenting for SARS-CoV-2 PCR testing during the first wave of COVID-19 in New Zealand. A is coloured according to the quality of representation of each symptom variable category (e.g. Cough and No-cough combined), which is the total of the squared cosine (cos2) values measuring the degree of association between variable categories and each axis, whereas B is coloured according to the contribution of each level within variable categories to the definition of the two dimensions (expressed as a %). C, D show the contribution of each of the 15 most important variables to the variation in dimensions 1 and 2. The red dashed line is the expected average value, assuming all contributions were uniform (i.e. the expected value under the null)