Generalised linear models (GLMs) allow to perform univariate statistics for NON-NORMALLY distributed data and they represent an alternative to performing general linear models on transformed data.
- This is a worked example by Gavin Simpson on how to analyse binary data such as survival data or presence/absence. He also shows how to compute confidence intervals correctly.
- On the difference between using GLMs or using LMs on transformed data, see here and here.
- On how to compute Confidence Intervals, see here and here.
- On how to calculate the SE of a quantity on the original scale knowing its SE on the transformed scale, see here (delta method).