Assessing Variable Importance for Best Subset Selection

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

See the attached pdf file.

Comments for author File: Comments.pdf

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

Review of

J. Seedorff and J. Cavanaugh   Assessing Variable Importance for Bet Subset Selection.

Some details:

 

Line 28 “it is well known” — give an appropriate reference

Lines 120-121  (A)  and (A∁)  —  A  and A∁,   as used in the mathematical expressions below.

Lines 132-134. Explain, it is not clear.

Lines 140-141 What is the rationale for that choice?

ρ ∈ {0, 0.9} is enough to reach conclusions?

Lines 172-173 “as the correlation between variables increases.” — there are only results for ρ ∈ {0, 0.9}, so the claim is excessive.

Line 191 — what is p here? The fact that in the Appendix p is used for the number of models or of variables, but so far the nonitalic symbol p had only been used in p-values.

Line 280 … p variables  ­— I think that  it would be better to use another symbol for the number of variables, the use of p-values and p variables and p models  is in my opinion inappropriate.

In the Figures, why rho and alpha instead of r and a?

 

General comments: The paper is well written, with judicious comments. From the fact that the presentation of results is restricted to ρ ∈ {0, 0.9}, some issues seem inconclusive. Figures should be commented with more details, since the deviation from the declared value of a is large, namely for large sample sizes.

Its subject seems to be very loosely related to the aims and scopes of Entropy, and so I hesitate to recommend its acceptation.

 

Comments for author File: Comments.docx

Author Response

Please see the attachment.

Author Response File: Author Response.pdf