Seulki Koo

I suppose The term "test gain" typically refers to the performance evaluation on an independent test set, which is distinct from the training set.

In your case, since you have set the test percentage to zero, it is indeed unusual to see a "test gain" value.

To clarify, the purpose of setting the test percentage to zero is to evaluate the model's performance on the training data itself, rather than on a separate test set. This is known as cross-validation, where the training data is divided into multiple subsets (in this case, replicates of 10) to assess the model's performance across different partitions of the data.

Regarding the use of the jackknife method, it is typically employed to assess the influence of individual variables on the model's performance. By systematically removing each variable and retraining the model, you can evaluate how much the exclusion of a particular variable affects the model's performance.

The "Jackknife of regularized training gain" likely indicates the performance of the model on the training data when each variable is removed one at a time. Similarly, the "Jackknife of test gain" could indicate the performance if the model were applied to the test data with each variable removed individually.

As for the use of bootstrap in this process, it is possible to employ it for resampling and assessing the stability of the model. However, whether it is appropriate or necessary depends on the specific context and objectives of your analysis.

Please recommend my reply if you find it useful , Thanks