@davidmanheim Sorry, guess I wasn’t clear. We’re not double-dipping—it’s a 3-way split. To predict, say, 1974, we first remove 74 from the dataset. We then fit a model using LOOCV to all OTHER years, excluding 74. The 1st fold excludes 74 & 42, the next excludes 74 & 44, etc. 1/n
-
-
-
We find the values of λ and the elastic-net mixing parameter α that minimize cross-validated error *within the dataset that excludes 1974*. Then we use the resulting model, which was not exposed to 1974 data for either training or testing, to predict 1974. We then repeat (2/n)
- 2 more replies
New conversation -
-
-
Lay off man, can't you see the tweet said <END>
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.