On 19 June 2020, I will be giving a talk at the eRum2020 conference titled “brquasi: Improved quasi-likelihood estimation”. The talk is about the brquasi R package, which provides a glm method for improving bias in quasi likelihood estimation.

That package is a spin-off of theoretical work with Nicola Lunardon on methods for reduced-bias M-estimation in general that require neither re-sampling (like the bootstrap does) nor full knowledge of the underlying distribution of the data (like other higher-order asymptotic methods do). The technical details behind reduced-bias M-estimation can be found at

Kosmidis I, Lunardon N (2020). Empirical bias-reducing adjustments to estimating functions. arXiv:2001.03786

You may also want to look at:

  1. the MEstimation Julia package that provides template M-estimation modelling and reduced-bias M-estimation using automatic differentiation!

  2. the Twitter thread here about reduced-bias M-estimation

Update (20 June 2020)

The slides from my eRum2020 talk are available here.