R Workshop for Social Scientist Discussion during Workshop
Please click on the title to see the session description
Tuesday, September 22, 2020
-
You can use them interchangeably, but it can be useful if you have a conjunction or some other text that uses ' or ' (e.g., 'can't' or 'as I quote 'blah' '). If you have both ' and ' in your text, then you need to use a regex format, which is a whole other can of beans.
To learn more about regex: https://regexr.com/
-
Nope! You can load and work with almost all kinds of data!
-
the use of '=' as an assignment operator dates back to Fortran and other early generation programming languages. Given that '=' has a predefined meaning in mathematics, I agree it's a bad choice of symbols for an assignment operator.
Further explanation here on the '=' and '<-' discussion: https://stackoverflow.com/questions/1741820/what-are-the-differences-between-and-assignment-operators-in-r
-
One thing is that there are functions that can parallel process operations on elements of the list. There are different more 'for' looking functions that can do can also parallel process things. Another difference is that any assignment in a for loop appears in the global environment if the operations were done in something like lapply they would not.
This is a great overview of the pros and cons of vectorization and looping: https://bookdown.org/rdpeng/rprogdatascience/vectorized-operations.html
-
I find that scripting a for loop takes about the same amount of time as lapply. If you have operations where subsequent iterations depend on the results of prior iterations, I find it's better to use a for loop.
And FYI, vectorization is also an option in matlab and Gauss, in case anyone uses those languages... https://bookdown.org/rdpeng/rprogdatascience/
-
https://www.cio.noaa.gov/noaa-style-guide/design-elements-left-nav.html
https://github.com/emilyhmarkowitz/NOAAFisheriesComms
https://github.com/James-Thorson-NOAA/VAST/issues/163
https://github.com/jordanwatson/NOAA_Branding/blob/master/NOAA_branding_508.R
Another package I use sometimes (and Jordan mentions above) for color palettes is RColorBrewer.
You can see some of the options here: https://www.r-graph-gallery.com/38-rcolorbrewers-palettes_files/figure-html/thecode-1.png and https://colorbrewer2.org/
Wednesday, September 23, 2020
-
[good question]
-
'Application programming interface' For more information about getting access to the Census Bureau's API: https://www.census.gov/data/developers.html. For more information youcan 'request a key' from there. https://github.com/sboysel/fredr, for pulling time series from st louis fed.
-
The costs come in at so many levels. Burden on the analyst is one, but information that can't be used by industry or in management also surely has a huge cost
Thursday, September 24, 2020
-
Both! So much more to talk about on that!
-
Good comments are essential for this. NWFSC (?) uses pretty rigorous code review, but it's very ad hoc - just checking things that occur to us at the time.
One idea: I've tried setting up companion excel files to record standard checks like min 10% check on some calcs or checks of data frame dimensions when filtering. Having the author set up what needs to be checked is helpful, but we didn't get very far with this. This was a practice I carried over from a private consulting firm, but we haven't gotten people in the habit in our branch (yet - fingers crossed).
One of the videos I put up on the Workshop website has a bit of a deep dive into the here() package.
R packaging (and adopting all the function documentation features) is a central part of my coding and quality. Perhaps a middle ground between adding another required layer of review for QA/QC would be to routinely include code in reproducible (e.g. RMarkdown) form with the deliverables for a project. Wouldn't mind reviewing code if it were as well organized as Emily's :-)
A related issue is when we have academic contractors do work, we need to require and follow through on requirements in contracts for code delivery.
Writing your code with the expectation that someone else will be reviewing it some day tends to make it better and forces you to do things like annotate, etc. A lot of how we handled the additional workload of QA/QC at my last company was by having newer folks who were trying to learn R do the QA. It made authors annotate well, and it allowed new learners to understand better what the code was doing.
-
A talk back in March that might give some insight on that if this is helpful. https://www.youtube.com/watch?v=-mycRwaC60A
-
[good question]
-
The upfront learning cost of LaTeX is much higher, and while you can perhaps do more with it pdfs than markdown. In most cases you can do enough with markdown to get done what needs to be done.
-
There is a lot that has been built into rmarkdown for 508, but some that is still lacking. I've had to look into some of that for the word-output markdowns was build for FEUS, but it also varies for html, pdf, other output formats. We've made the pdf 508 compliant after the pdf was generated, which is a pain. I tried to look into it once a while back but at the time I didn't find a good way to automate 508 compliance in LaTeX.
-
I have a section on getting Git/GitHub working with RStudio in a workshop I taught over the summer. Geared to those who have never used Git or used it and got very frustrated. https://rverse-tutorials.github.io/RWorkflow-NWFSC-2020/
Also has info on working with internal GitLab repos, if you are working on confidential material that must stay on a NOAA server. (eli.holmes@noaa.gov)
-
Link to Hadley Wickham's e-book on package development: https://r-pkgs.org/
-
Nice info about reprexes: https://github.com/tidyverse/reprex