Attributable and unattributable risks and fractions and other scenario comparisons

Scenarios can be defined as alternative versions of the same dataset, with the same variables but different observations and/or values. Applied scientists frequently want to predict how much good an intervention will do, by comparing outcomes from the same model betwen different scenarios. Alternatively, they may want to compare outcomes between different models applied to the same scenario, as when standardizing statistics from different subpopulations to a common gender and age distribution. Standard Stata tools for scenario means and comparisons are margins and pwcompare. A suite of packages is presented for estimating scenario means and comparisons using margins, together with Normalizing and variance–stabilizing transformations, implemented using nlcom. margprev estimates marginal prevalences, marglmean estimates marginal arithmetic means, regpar estimates the difference between 2 marginal prevalences (the population attributable risk or PAR), punaf estimates the ratio between 2 marginal arithmetic means (the population unattributable fraction or PUF), and punafcc estimates a marginal mean between–scenario risk or hazard ratio for case–control or survival data, also known as a PUF. The PUF and its confidence limits are subtracted from 1 to estimate the population attributable fraction (PAF). Formulas and examples are presented, including an example from the Global Allergy and Asthma European Network (GALEN).