The Royal Statistical Society is pleased to present for reading and online discussion the paper ‘Assumption-lean inference for generalised linear model parameters‘ by Stijn Vansteelandt and Oliver Dukes, Ghent University, Belgium. The discussion meeting will take place on Tuesday 6 July 2021, 2-4 pm (BST) with a DeMO (pre-meeting) with the authors at 12:30 to 1:30 pm on the same day. The paper’s abstract is produced below:
Inference for the parameters indexing generalised linear models is routinely based on the assumption that the model is correct and a priori specified. This is unsatisfactory because the chosen model is usually the result of a data-adaptive model selection process, which may induce excess uncertainty that is not usually acknowledged. Moreover, the assumptions encoded in the chosen model rarely represent some a priori known, ground truth, making standard inferences prone to bias, but also failing to give a pure reflection of the information that is contained in the data. Inspired by developments on assumption-free inference for so-called projection parameters, we here propose novel nonparametric definitions of main effect estimands and effect modification estimands. These reduce to standard main effect and effect modification parameters in generalised linear models when these models are correctly specified, but have the advantage that they continue to capture respectively the (conditional) association between two variables, or the degree to which two variables interact in their association with outcome, even when these models are misspecified. We achieve an assumption-lean inference for these estimands on the basis of their efficient influence function under the nonparametric model while invoking flexible data-adaptive (eg machine learning) procedures.
The full preprint is avaliable to read here.
The authors will present their paper in an online meeting that we very much hope you will join. We welcome contributions to the discussion following the presentation of the paper. Please let Judith Shorten (j.shorten@rss.org.uk) know in advance if you would like to reserve a speaking slot (up to 5 minutes) so that your name can be called.
Anyone can comment in the open session that follows; we’ll do our best to fit everyone into the time allowed. All written versions of oral contributions will be published with the paper in Series B of the journal (see more below).
Register for the Discussion meeting
Register for the DeMO (pre-meeting)
Written contributions:
We welcome contributions to the discussion in writing for publication in Series B. Your comment must be no more than 400 words in length (not including tables, reference details and figures). Please add the following information to the beginning of your contribution:
- the author(s) and title of the paper under discussion
- your name, title, affiliation
- an email address for receiving the proofs to check (these will be issued about 3 months after the meeting).
We encourage everyone to comment at the meeting in person. However, written contributions received before 6 July will be read at the meeting (if time allows) for anyone unable to join.
Submit your written contributions in an editable format (word doc or tex). Please supply a pdf copy for review and include a separate file for each figure. Send direct to j.shorten@rss.org.uk before Tuesday 20 July 2021. This deadline is important to help the authors of the paper, who will consider all the contributions and compose a reply in limited time, and to help the timely production of the journal. A condition of publication of your contribution is that you assign copyright to the Society on the form that will be provided when we send you the proofs for checking