Statistical modelling

Human infections

We first exclude Site 1 using Spread → Restrict/Filter → To Groups (factor levels)... and then analyse the variable P_infected. The data are balanced with three age groups, male and female in each of four sites so we could analyse the data using a randomised block model. However, we notice that separate values of P_infected are based on different numbers of individuals; this influences slightly the precision with which each P_infected value is determined. We shall, therefore, show how the data can be analysed using a weighted analysis of variance.

We start by applying Stats → Regression Analysis → Generalized Linear Models... and including just age in the model. By clicking Options... and clicking 'Accumulated' and entering N_age_gen into the 'Weights: box' (see dialog box alongside) we obtain an analysis of variance for age alone. We can then click Change Model in the original dialog box and add gender to the model, and then repeat and add the interaction.

From the resulting analysis of variance we see that both age and gender are significant, but not the interaction. We can then click Change Model again and drop Age.Gender to return to the model with just the two parameters: age and gender.