Closed GaryBAYLOR closed 1 year ago
I'll add some assorted comments here, in addition to the above:
in most cases
) that I think weakens the presentation overall. that can be addressed later, and should involve specific examplesasymptotically
here is important: how can we evaluate if our data and models are appropriate for this package, or that we can expect reasonable convergence?In terms of the review checklist, I think the two points description [...] for a diverse, non-specialist audience
and clearly states [...] who the target audience is
could both use further elaboration and expansion. The author is very clearly knowledgeable about the methodology in use here. I think the target audience would likely be users of standard marginal models (e.g. GLM), but observe that many such users are specifically using GLMs because they understand basic regression models but are perhaps not as comfortable with more advanced ML techniques. I think including some content that specifically reaches out to such users, if indeed they are part of the intended user set, would be most welcome.
@GaryBAYLOR I have made edits to the paper based on your comments. @cpalmer718 I have done some rewriting to reference existing methods and lay a quick but better foundation as to the methodological gap this method/software is filling. I've added some additional details as to asymptotic convergence and under what assumptions we have an efficient rate of convergence. I have also rewritten sections of the background.
@blind-contours Looks good. Just one very minor comment, be consistent in writing nonparametric
, either non-parameter
or nonparametric
is okay (line 18, 19, 30, 44)
@cpalmer718 I'm good on my comments. Please feel free to close the issue when the questions in your comments are addressed.
I've updated the md document to be consistent in using non-parametric.
line 10:
glm --> GLMline 30:
Thus --> Thereforeline 39
: glm's --> GLMsline 48, line 50
: semiparametric --> semi-parametricline 58
: unbold "a priori" and change it to italic formatline 69
: average treatment effect --> average treatment effect (ATE). The reason is, when using an acronym, state it when it first appears in the paper and only use the acronym through the rest of the paper.line 84
: average treatment effect (ATE) --> ATEline 87-88
: Remove 'language and environment for statistical computing'line 91
: CVtreeMLE --> Itline 92
: sl3package --> sl3 packageline 95-96
: the hyperlink to CRAN is invalid as the package hasn't been published there.line 102-103
: To make sure the title of the reference matches with the title in DOI, sl3: Modern pipelines for machine learning and Super Learning --> sl3: Modern super learning with pipelinesline 105-106
: Statistical Inference for Data Adaptive Target Parameters --> Statistical inference for data adaptive target parameters. This is to make all the references look consistent.