In vanheeringen-lab/learnr.proto: A set of LearnR Tutorials.

knitr::opts_chunk$set(echo = TRUE)

Introduction

In this rmd I wrote down

• Things I/we learned along the way (and could be applied more in the beginning).
• Ideas for other questions (we had or I had but left out because of time-constrains).
• Anything else that might be worth to tell the next person working on this.

Things I/we learned along the way

chunk headers

Problem: Hints did not always display properly.
Solution: Turned out TRUE should be written as 'TRUE' in chunk headings when working with learnr. If you use for example exercise=T, the hints will not be displayed (annoyingly).

coding exercises

Problem: For fill-in-the-code exercises, we add '___' to the exercise chunk but this gives an error when 'Run Document' is done.
Solution: You should add 'error=TRUE, exercise.eval=FALSE' to the exercise chunk headings.

giving feedback

Checking answers is done with gradethis package. There are (at least, as far as I used) two options for answer checking:

1. grade_result()

2. use the chunk title [exercise_chunk_name]-check when you use grade_result().

3. formulate as, for example:

grade_result(pass_if(~identical(.result, head(distance2NearestFeature(monocytes_h3k4me3, tss_chr19)))))

• You can use additional argument correct = [text] to customize the message displayed when the submitted answer is correct.
• In fg3 I realized that we can also compare to plots, eg:

grade_result(
 pass_if(~identical(.result, heatMatrix(mat = scores_h3k4me3_tss, 
           xcoord = c(-5000, 5000), 
           col = h3k4me3_cols, 
           main = "Unordered H3K4me3 signal, centered at TSS", 
           xlab = "bp (TSS = 0)")
           )),
 correct = 
)

1. grade_code( [compare the answer with this answer])

2. use the chunk title [exercise_chunk_name]-code-check when you use grade_code().

3. provide the code solution in the chunk with title [exercise_chunk_name]-solution
4. this option works towards the exact code given in the solution chunk with nice feedback. BUT for the answer to be correct the code should really be exactly the same as the solution. Eg: head(a, 2) =/= head(a, n=2).
5. another BUT; when you want students to generate the plot and you use this method of grading, the plot won't be visualized. Work-a-round1: direct students to hit "Run Code" explicitely before or after "Submit Answer". Work-a-round2: use grade_result() instead.
6. another BUT not handy if you multiple lines of code and leave blanks. In that case, students might fill out one line and hit "Submit" but will receive weird, not-understandable feedback due to the blanks in the other lines of not-yet-corrected code. Work-a-round1: direct students to fill in all blanks before hitting "Submit Answer". Work-a-round2: use grade_result() instead.
7. preferentially leave out hints for this type of answer checking.
8. however, I still added hints to grade_code questions when there were multiple functions where they had to complete the code. This is not really optimal use of learnr. An alternative would have been to chop these questions up in smaller parts but then I would have to prepare setup chunks for each invidual, chopped-up-exercise (as I understood it).

{r naam-solution} exact de code die je verwacht

{r naam-code-check} grade_code()

Possible corrections for next year

• restructure fg2, took more time than fg3
• remove (code check) comment from exercises

Anything else that might be worth to tell the next person working on this.

• In prepdata.Rmd I wrote out commands I used to prepare some of the objects. Some of the raw data are too bit to include in this tutorial. They can (as for now) be found on cn45/scratch/ctoenhake/edu. --> now relocated to /molbiol/mb03-sdb1/mbdata/ctoenhake/edu/

• All BLUEPRINT data were downloaded from the blueprint data portal.

• I limited this tutorial to monocyte data of sample C000S5.

• Sometimes running learnr.dashboard:::.setup_learnr_proto(lib="/home/ctoenhake/learnr.proto/renv/library/R-3.6/x86_64-pc-linux-gnu", pkg="/home/ctoenhake/learnr.proto") gives at the end the error:

 DONE (learnr.proto)
Error in validate_signal_args(.subclass) : 
  could not find function "validate_signal_args"

+ solve by restarting R session

• Sometimes running learnr.dashboard::start_tutorial("fg3") after you've installed the new/corrected tutorial in the package with learnr.dashboard:::.setup_learnr_proto(), hint chunks with empty functions give an error, eg:

label: q3b_h3k4me3plot-hint-3
Quitting from lines 554-556 (fg3.Rmd) 
Error in parse(text = x, srcfile = src) : <text>:2:1: unexpected input
1: # eg
2: _
   ^

--> solve by restarting r and run learnr.dashboard::start_tutorial("fg3") again.

• If you run learnr.dashboard:::.setup_learnr_proto(lib="/home/ctoenhake/learnr.proto/renv/library/R-3.6/x86_64-pc-linux-gnu", pkg="/home/ctoenhake/learnr.proto") and there is a fg#.html in the tutorials dir. Than that html will be loaded upon running learnr.dashboard::start_tutorial("fg3") instead of your updated tutorial. solve by removing the .html before installing the tutorial.

vanheeringen-lab/learnr.proto documentation built on March 1, 2021, 11:10 p.m.