Luke Sonnet

We can redeem this issue, see the title change

This can definitely be improved, but currently it would return no coefficients, so it would be a bunch of infrastructure to return a bunch of empty objects. Do we want...

Decision: can I just have a smarter error? I don't really want to support this.

lm_robust assigns a colname via `model.matrix`, so even if you pass it a matrix with no names it works. It's just lm_robust_fit, and you're correct I just need to set...

(b) is implemented in 8886b94, leaving this issue open and tagging as future to figure this out.

The above issue #2105 has a good example of how the status quo manifests in unclear errors. I've also written more about this in a community PR [here](https://github.com/growthbook/growthbook/pull/1962#pullrequestreview-1816300423), copied below:...

I think it's more likely we support a variety of time series analyses which help you understand how effects change as you roll up more days after exposure. E.g. a...

One other solution is to keep the First Exposure model in effect but use a Fact Metric where you can turn conversion windows off altogether to create two versions of...

Unfortunately yes, that is what you'd need to do. We're considering making these kinds of "metric variants" easy to create, manage, and add to an experiment, and we're also actively...

The metrics page in general lacks a lot of functionality to make it a full-featured insights tool. Right now it serves as more of check as you set up metrics...