Background: I am running a univariate linear regression model (OLS) and outputting a prediction interval for Y_hat. The linear model mostly suits my task up to now as it provides needed interpretability, while being simple and fast.
Now, however, more predictors are going to enter in the model (still it would hold that n > p) and since some of those might be highly correlated, I am looking into using a penalized linear model while still being able to output a prediction interval for Y_hat.
The set of my predictors is NOT big, (less than 15-20 for around 300 observations).
Question: I realize that prediction intervals for ridge/elastic net/LASSO are not straightforward as in OLS since the estimates are biased (as per posts e.g. https://stats.stackexchange.com/questions/121993/elastic-net-confidence-intervals-for-parameters and page 18 here: https://cran.r-project.org/web/packages/penalized/vignettes/penalized.pdf)
I was wondering however if there are some already implemented (at least partial) solutions (Python packages particulary) that I can use? If not that, at least some general guidelines on how one can handle such types of problems would be highly appreciated.
Thank you
