francisco-orejarena
commented
6 years ago I regret leaving the anova comment in there. I remember writing it, but when I tried to run the function, the only model that worked was the duration, and I felt that solely having one anova function run was a bit disingenuous I guess? Either way, I didn't feel that I could interpret it well enough, but it was a mistake to leave it in the paper. The code in regards to language numbers doesn't exist I don't believe. I might have phrased that poorly, but it was essentially me going into the excel document and adding a new column with a number for the language. A lot of small mistakes in this project definitely built up, but I'm glad you wrote this feedback. I think I could definitely make it stronger with another iterations, and maybe a stronger grasp on stats, as sometimes I fell short. Thanks Chris!
Comments: This project builds on an earlier study which evaluates the FLH, and takes a different statistical approach. The introduction is strong and clearly states the main purpose of the paper.
The author describes the steps in reorganizing and tidying the original dataset, providing good justification for any variable exclusions, and includes examples of the resulting data frames.
Using a binomial regression model to analyze the data was appropriate for this project. The author stated which models he fit, and provided full summaries of these models in the text of the paper before interpreting the results. This is a clear and transparent way of presenting the findings, and is a strong point in favor of this project being considered reproducible.
Questions: Where are the results of the likelihood ratio test in your paper (using the anova() function)? I found the test in your code, and you mention it in the introduction, but I was unable to find it in the Results section.
Also, where in your code did you number the languages? I saw where you renamed the ‘lang’ column to X1, but couldn’t find where you changed the language names into numbers. How did you do that?
Suggestions: Since your paper is an improvement on the statistical methods used in Lunden et al’s analysis, I think it would be helpful to provide more detail about what they did in their paper. You allude to comparative models in the introduction, but I’m not sure about the details.
Building on that, I think that in your discussion/conclusion section, it would be good to explain why your models are an improvement on the original analysis. You come to the similar conclusions with respect to the validity of the FLH, but what is it about your model that justifies this result more clearly/intuitively/etc. than Lunden? I think this would give the conclusion a stronger punch.