3.2.1. Log-transformation

Exponentially distributed data (i.e. quantitative virus titre data) must be treated with a power transformation (Box and Cox, 1964; Bickle and Doksum, 1981), usually a log-transformation, before they can be used to estimate descriptive statistics (means, medians, variances etc.) or be used in parametric statistical analyses (ANOVAs, correlations, GLMs etc.). This can be done prior to statistical analysis, or can be incorporated as part of the statistical analysis.