Consider warning about the GI

[kaitejohnson]

If you want users to fit to their own data in user testing, you might want to consider some warning message about the GI being arbitrary and that it should be user-specified.

For example, if someone was following the vignette, they might think they could just take their favorite respiratory virus case data and run this one line to get the R(t)

cfa-gam-rt/vignettes/RtGam.Rmd

Line 73 in ce33553

fit <- RtGam(

(that someone being me in user testing).

I would also suggest being a bit more clear here that fitGam is learning the trend in cases, and that not until the predict /plot function will you get R(t)

[zsusswein]

What do you think @seabbs? I think we likely need a good "formatting PMFs" vignette here and/or in {primarycensored} to make things clearer, but there's a bunch of higher priory stuff to get to.

We're pretty clear in the vignette and the docs that this is a simulated example from an SIR model:

cfa-gam-rt/vignettes/RtGam.Rmd

Lines 125 to 126 in ce33553

	In this simulated example, cases are observed immediately so the mean delay is 0.
	We have pre-generated the generation interval probability mass function (see more with `?sir_gt_pmf`).

Perhaps a big red warning box would be good for clarity? @kaitejohnson's point that calling RtGam() perhaps does imply that we're going to give the user an $R_t$. That said, I don't love the flow of requiring a GI at the time of fitting the model if it's not strictly required.

[seabbs]

I think in the vignette I would have a sub section with a plot where you introduce the generation time and the delay as a clear other thing (something like this https://package.epinowcast.org/articles/single-timeseries-rt-estimation.html#specifying-the-model-using-epinowcastenw_expectation) would help.

RtGam

I think this point about the name is a really really strong one and does suggest to me it should maybe be called something else (unless we do plan to add the offset based renewal so you would actually get Rt). Perhaps that means calling the whole package something else...

I don't know about that but I definitely think its worth thinking through clearly at this stage.

One of the reasons for moving to a name like EpiAware vs my earlier also great names was trying to break the link to nowcasting, EpiNow2 doesn't have Rt in the name either because the idea was it was all about everything you need to know now (but that got a bit linked with nowcasting in epinowcast).

[seabbs]

If we did go this way here are a few LLM suggestions to get you thinking:

• GrowthView
• SmoothTrend
• FlexModel
• GrowthGam
• FlexTrend
• SmoothView
• GrowthLens
• TrendModel
• FlexView
• TrendWatch
• GrowthTrack
• TrendScope
• SmoothModel
• TrendGam
• ModelView
• TrendSight
• GrowthScope
• SmoothTrack

"FlexModel" or "SmoothModel" might be most appropriate as they:

• Reference the flexible additive modeling approach
• Avoid being too specific about methodology
• Maintain professional terminology
• Are clear but not overly technical
• Leave room for future methodological additions