PhD Meta

Data visualization can be an effective way to get the 3 or 4 messages of your manuscript across very quickly. Owing to this though, data visualizations can be very tough to do well, and you can spend a lot of time messing with the small details.

In general, try and keep things simple - what is the easiest and most convincing way to make your point?

Tables are really tough - when you have a table in mind, think in columns (most journals have 2-3 columns per page), and look at what the journal does in their previous publications

For figures, some guidance from Dr. Matthew Kay at the University of Michigan:

• Locality
• Order
• Layering
• Grids and synchronized axes
• Obey the pen
• Treat visual attributes like adjectives
• Color

For plotting in R:

• Color palettes for various needs, including black/white printing and color blindness: link
• Another color palette called visibly
• Color Brewer: discrete color palettes
• The R Graph Gallery: thanks MS!
• Colorblindness testing: https://colororacle.org/

Over the time of my PhD there were some resources that I consistently returned for stats guidance:

• A table for help choosing the right statistical test for your data: UCLA IDRE
• Introduction to Data Mining, Stanford Stats 202: Intro to Statistical Learning

Friends have also pointed me in the direction of other helpful resources:

• Intro to Bayesian statistics
• Intro to Generalized additive models: link
• Mixed models with R: link 

Just like the REI triathlon checklist, but for manuscripts!

First, think you're done? Put it down for a few days, then try to get to < 5,000 words, max 6 tables/figures, max 50 refs (see JESEE) 

There is also something called the EHP Section 509 guidelines for figures?  https://ehp.niehs.nih.gov/authors/figures. JESEE also has guidance for figures that specify specific max dimensions

Then:

• Use a 3rd party tool to check: https://writersdiet.com/ 
• get rid of idioms (e.g., "tailoring")
• if you upload figs as docs, probably helpful to include captions in docs
• make sure supplemental information has header (title authors etc)
• check for typos in figures and tables
• make sure all your figures have the same text font (serif etc)
• Look at the journal guidelines for new submissions
• First, proof it well, then set it down for 3 days, then print it out and read it again. 
• Get rid of as many "e.g.," and "for example" and "As well" as you can
• use "draft print" mode, print 2 sheets per page, double sided, grayscale
• Have you spelled cover letter names correctly?
• Does the paper header have correct contact information?
• Have you added line numbers? (Are they referenced correctly in response to reviewers?)
• Tables and figures (think in columns) – check that all are referenced correctly
• Also think about colors, if you have a color background to an image, what will that look like if its is printed in black and white?
• Are all references formatted correctly / are correct? change your reference formatting to the journal's so you can see what it will look like? No weird characters or formatting errors or abbreviated journal names? get rid of et.al,
• Do a spellcheck / grammar check?
• numbers less than ten are written as words
• Are Methods/Results in past tense?
• All abbreviations/acronyms defined?
• Are all subscripts correct (e.g., PM_2.5 not PM2.5)
• All the same significant figures? and does the simple math of columns in tables line up (e.g., 0.1 * 0.2 = 0.2 and not something else because of rounding)
• All figure #s and section #s correctly referenced
• Double check email from journal (e.g., should all changes be highlighted etc)
• Get rid of all double spaces after periods
• Read the manuscript backwards - you spend a lot time reading the beginning!
• Double check all your numbers, that they are replicated in the tables etc

For many reasons, Endnote is a great tool. However, when it breaks, it can be a pain to go through a find all of the your references, so you can replace with something like Mendeley. I recently had to go through this process, and here is what I learned:

1. To remove all hyperlinks in a Word document, press CTRL+A to select the entire document and then press CTRL+SHIFT+F9. 
2. Then do a find replace using the following (source):
• FIND: (\[)(*)(\])
• REPLACE: \1\2\3
• And if you format the response with 'Highlight', then you can easily find your new unlinked intext references!

• You want to be short and to the point, just to respond
• Want an end to the discussion
• Answer the question
• Make it easy for the editor to see that we’ve answered the reviewers questions
• Don’t send in response to reviewers late, its not good practice, risks getting reviewed as a new paper
• Don’t do 100% perfection work, don’t let the perfect stand in the way of the good

Response. We thank the reviewer for their comments and insights and have revised the paper and undertaken some additional analysis in response.

Here is an example of the formatting of a recent response to reviewers I was working on: