Option to manually select triangle grain

[LegendaryPie]

Description

I'm using models that have been trained on OMDM data and am applying it to many different triangles of sparse data. And it is coming up that sometimes all of the accident periods line up in the same month of the year or line up in quarters and that changes the grain due to the inference nature of it. Which then causes errors when I use the model to predict on that triangle due to them being different grains.

Is your feature request at odds with the scope of the package?

• Yes, absolutely!
• No, but it's still worth discussing.
• N/A (this request is not a codebase enhancement).

Describe the solution you'd like, or your current workaround.

I would like the ability to specify in triangle creation the grain that I want the triangle to be so that I can have consistency.

Do you have any additional supporting notes?

No response

[henrydingliu]

The grain inference logic is doing some heavy lifting in preventing nonsensical triangles. If we were to add grain selection to the triangle declaration, the most sensible implementation would actually be to allow the inference to go first, then apply a grain change after. My suggestion is to treat this as another grain bug (e.g. #609). So if you can provide some examples of where you can't use the grain method to achieve your desired grain, we can try to figure out why the grain inference logic is not doing the right thing.

[LegendaryPie]

Changing the grain afterwards could work fine, but the inference is setting it to OYDM or OQDM and the grain I want it on is OMDM which doesn't work with setting the grain since it is going to a finer grain.

[henrydingliu]

understood. the inference isn't working cleanly. can you provide a concrete example?