Rémi Thériault

@bwiernik I'm planning on submitting 443 before this is resolved, but just pinging you in case you have some opinion on this that I could integrate before the merge.

All four questions above. But we can start small, with no 1. (You can skip the reprex, it was only a demonstration of the point)

Thanks. For reference, here are the other thresholds in case you'd like me to change any other one: ```r list( zscore = stats::qnorm(p = 1 - 0.025), zscore_robust = stats::qnorm(p...

Ok I have resolved no 3. It is because the threshold is not used _on_ the distance score as in other methods, but in an _earlier_ step to _calculate_ the...

About no 4: I think it is not a bug. I will make a reprex shortly to show that it is calculated like this on purpose. However, it does beg...

Here is the way IQR distances are calculated. Is this a bug or intentional, and should we change it? ``` r x 6 6 0 0 0 0 0 0...

Would it make sense to report the actual percentile as the IQR distance score, regardless of whether `method = tukey` or not, since only the cut-off values would change in...

@bwiernik Would it make statistical sense to report the actual percentile as the IQR distance score? What about the CI distance score?

Ok so if everyone is alright with that, I'm going to submit a PR with all our agreed changes, including changing the distance scores to percentiles for IQR and all...

> I'm currently inclined to have a simplified function for lm models only, using the model's matrix (instead of formula interface), which will allow for interaction support. The current version...