Reverse a Yeo-Johnson transformation

[conarr5]

I have a project where I'm using step_yeoJohnson() to transform my data and then adding noise to the transformed data. How can I use the recipes package to reverse the YJ transformation?

[topepo]

Can you give more information about what you would like to do?

All of the operations in a recipe occur before the supervised model. I'm not sure how, in a recipe, un-doing a transformation helps. Here's a high-level sketch of the process:

1. start recipe
2. transform data
3. do some other things using the transformed data
4. back-transform data
5. do some additional things using the untransformed data
6. pass data to model (untransformed)
7. make predictions

Can you help us understand what things would go into steps 3 and 5?