Ideas for next year

[vincenzocoia]

Starting a thread to put ideas for next year.

Tibble Joins lecture

• Would have been useful to demonstrate what happens when we have a song like "What's love got to do with it" by Tina Turner in one tibble, and the same song by another author in another tibble.
• Would have been useful to demonstrate joining a tibble with two predictions, where the columns have the same name in both tibbles, showing that you don't always need to choose columns that are in common between the two. Something like:

  set.seed(1)
  x <- rnorm(5)
  model1 <- tibble(x = x, yhat = 2.1 + 3.2 * x)
  model2 <- tibble(x = x, yhat = 1.5 + 2.9 * x)
  left_join(model1, model2, by = "x")

  And, perhaps follow with a pivot_longer() -- as an alternative to the clunky bind_cols() then bind_rows() method.

General Idea

For the tibble joins lecture today, I did a little segue into a higher-level discussion about where tibble joins often show up in practice. I think it was useful for framing the topic, but it also brought up other aspects of writing a data analysis that's hard to treat as its own topic, such as using mass[1] vs unique(mass) to retrieve island land mass.

[vincenzocoia]

here::here() lecture

Definitely important to go over the notion of running code interactively vs. from source. Too hard to explain this at the same time as trying to explain why here::here() sometimes gives different things between the two.

• Rmd: interactive = just execute the cells; from source = knit.
• R script: interactive = just execute code (moves it to console); from source = "render report" or source().

EDIT: I should avoid this topic, because it is discussed later in 545B in the "Automation" topic.

[vincenzocoia]

Worksheet 1B

I think there might be too much emphasis on having functions print() things as opposed to just outputting a character vector.

[vincenzocoia]

Mini Data Analysis project

Instead of allowing them to work with any dataset, choose 3 that will definitely work nicely with the questions posed in the project.

If you are invested in a dataset of your own, encourage them to use it, but also describe it so that the teaching team can understand it. Perhaps even get them to put it in a data folder with a README describing each column -- or at least, the most important ones.

[vincenzocoia]

Mini data analysis milestone 2

Questions ended up being too specific -- would be more effective to generalise. This milestone was also too demanding for students.

[vincenzocoia]

Mini data anlysis milestone 3

People were copying their knitted md files into the output folder. Indicate not to do this (it's not reproducible anyway)

[dy-lin]

Feedback for Assignment 1-B in the private instructor repo: https://github.com/UBC-STAT/stat-545-instructor/issues/24

[vincenzocoia]

Assignment 1-B

• [x] RRR not clearly specified.
• [x] Submission instructions should link to tag release, not the md file (as specified).
• [ ] Should include an example description (like the text that appears below the title).
• [x] English in description should be stricter (grammatically)
• [x] verbose option should be stricter. Shouldn't be square(): "calculating square..." "...done calculating square!". Should be a multi-calculation function.
• [ ] Lots of instructions here. It would be better to put all the instructions in the tables at the start of the section.
• [x] Separate checklist document is more useful than embedded.
• [x] Should be no penalty for simplicity of functions -- that will be naturally taken care of in the downstream questions.
• [x] Exercise 2 -- everyone did very well, because I went over this in class.
• [x] Bonus (contribute to an R package) -- should be part of Assignment 5B, which itself should be like a project for the whole course.
• [x] Should include a basic resource for how to write a function in R, because I don't go over that so much in the class.

[vincenzocoia]

Policies

• [x] Late assignments should not have 24h limit to receive the 50% penalty because there's no "solutions" document that needs releasing.
• [ ] English, spelling, and grammar: if you have issues with this (eg. maybe you are learning English during the course, or maybe you have a disability)

[vincenzocoia]

Assignment 2-B

• Let students and TA's know that we can check vignettes and install the tagged release version using these two arguments of install_github():

install_github(build_vignettes = TRUE, ref = "0.1.0")

• Need to actually say that we want 3 non-redundant tests, even though we already requested that in Assignment 1-B. Also, should require students to have 3 tests per function, in the case that they write more than one function.
  • Even though there were grades set aside for this in 1-B, there perhaps should be some grades set aside for quality of tests (and functions) -- in case they make new functions/tests, to encourage them to incorporate feedback, and to get redundancy built in.
• There's a note that says to briefly document -- but should remove "brief", because it's often necessary to elaborate.

[vincenzocoia]

General assignments

• Tidy submission: Would be useful to get students to provide a link on the diff between the template commit and the final commit -- useful for grading things like the BC Liquor app.
• Make assignments out of 100, because we ended up working with units of 0.25 when grading.
• Tidy submission: code should be styled appropriately -- tidyverse style.

[vincenzocoia]

General topic flow

At least for 545B, where topics span 2 days, it may be useful to cut out some instruction on more details, and end with a 10-20 minute presentation on expanding the topic. This would allow me to at least plant the seed for topics that I wish we could cover in this course, like Rmd presentations, bookdown, GitHub pages, etc.

[dy-lin]

Assignment 4-B: Makefiles

Using /usr/share/dict/words as a dependency for words.txt is causing issues for Windows users. Looks like this has consistently been a problem in the past years. I advised the current students to adapt this issue into code, and download words.txt from here:

words.txt:
    Rscript -e "code to download words.txt here"

It probably doesn't even need to be R code-- I'd accept using wget or curl for the students savvy with the command-line too, although it is good exposure to RCurl.

UPDATE: I just checked this year's Makefile, and it looks like there was a commented out line for them Windows users. Maybe we could write extra comments in the Makefile to let the students know to use that line instead if they're on a Windows machine.

[vincenzocoia]

Assignment 4-B: Makefiles

This assignment is not Windows-friendly. Should teach remake instead of make, and just allude to make at the end.

[vincenzocoia]

Assignment 3-B:

• Have them make this in a separate repository (doesn't make sense for the makefile analysis).
• Students are tending not to use interactive UI -- for example, filtering by subtype in the BC Liquor.
• It would be a good idea to have a peer review option for students to get feedback on their shiny apps somehow.

[vincenzocoia]

Assignments 1 and 4

• This should be a "data analysis" repo.
• Make sure they keep updating their main README in addition to the assignment README.

[vincenzocoia]

Assignment 5-B

strings assignment: there was an article that many people drew on for removing stopwords. It's OK to draw from it, but many people didn't cite it. I should use this as an example of how and when to cite code.

List the Gutenberg project as being a source of freely available books: https://dev.gutenberg.org/

[vincenzocoia]

Overall

Extensions:

• As for an extension only within 48 hours of the launch (this means it is a foreseen circumstance).
• Deadline no more than 1 week after the original deadline.
• Standardized way of asking for extensions -- google form?
• Should go over the policies section on Class 1.

Assignment feedback: there should be a document outlining procedures:

• How to interact with your feedback.
• Watch your repo.
• Dispute only 1 week after the grade (can't add this to the calendar, because this is a variable date. Instead: add a section on the assignment (akin to the "how to submit your assignment") labeled "how to follow up with the assignment".
• Don't use Slack for

Improving the stat545 content:

• The guidebook was missed this year. Would be good to make a hybrid guidebook: embed videos, etc.

Others:

• Importance of why we use github. Emphasize that we are using git and GH throughout the course (this was a big surprise for many coming in to the course).

Team leading:

• Vision for the course: indicate to the team and the students

[dy-lin]

Clarification on Assignment 4:

• a lot of students thought 'add dependencies' for each assignment in the README meant to list the file dependencies (i.e. dependencies for the target in the Makefile), instead of the 'package dependencies'
• some students tried to rm phony targets in their clean, so this may also need some more clarification

Overall clarifications:

• even for the later assignments, I still saw some students submitting local URLs or paths to their assignment instead of the GitHub link, e.g. /var/temp/private/etc or http://127.0.0.1:XXXX

[dy-lin]

Overall, I think there was quite a bit of 'hand-holding' in 545A, with the checklists etc. It shouldn't be our responsibility if the students miss certain criteria for not reading carefully. I also think that the students should be able to just download our .Rmd files for assignments, and not need us to create a pull request for each assignment. Although if this is the direction we want to continue on, and if the course coordinator is git-savvy, this could be one of their consistent tasks.

[vincenzocoia]

Here's a fun dataset that would be useful to introduce dplyr and ggplot2 -- an analysis of pride parade entries across time: https://github.com/GaytaScience/PrideParades

[vincenzocoia]

Yulia recommends ProjectTemplate as an alternative topic to Makefiles.