avehtari / bda_course_aalto Goto Github PK

The tasks in the assignment description and rubric questions have different order, which is confusing. Right now:
Q4 -- corresponds to task 2;
Q5 -- task 3;
Q6 -- task 5;
Q7 -- task 2 (and the "k-values should be visualized or tabulated" part from the rubric should probably be in the task description and not in the "Hint" section, since it seems like a requirement);
Q8 -- task 4;

So the right order should probably be:
Q4, Q7, Q5, Q8 and Q6.

Add to assignments page examples:

Here are some fictional examples of good help requests:

Chat: TODO: add BDA course specific example

Zoom: TODO: add BDA course specific example

For comparison, see Parallel computing course examples http://ppc.cs.aalto.fi/2020/lab/

I just tried to render assignment4.qmd but it failed due to a missing file.

Here is what I did:

[Code] git clone [email protected]:avehtari/BDA_course_Aalto.git
Cloning into 'BDA_course_Aalto'...
remote: Enumerating objects: 6270, done.
remote: Counting objects: 100% (190/190), done.
remote: Compressing objects: 100% (106/106), done.
remote: Total 6270 (delta 86), reused 181 (delta 83), pack-reused 6080
Receiving objects: 100% (6270/6270), 120.16 MiB | 5.63 MiB/s, done.
Resolving deltas: 100% (4009/4009), done.
[Code]
[Code] cd BDA_course_Aalto
[BDA_course_Aalto] git checkout gh-pages                                                master
branch 'gh-pages' set up to track 'origin/gh-pages'.
Switched to a new branch 'gh-pages'
[BDA_course_Aalto] cd assignments                                                     gh-pages
[assignments] ls                                                                      gh-pages
README.md               assignment4.Rnw         assignment8.pdf         simple_template3.qmd
additional_files        assignment4.html        assignment8_rubric.html simple_template4.html
assignment1.Rnw         assignment4.pdf         assignment8_rubric.md   simple_template4.pdf
assignment1.html        assignment4.qmd         assignment9.Rnw         simple_template4.qmd
assignment1.pdf         assignment4_rubric.html assignment9.pdf         site_libs
assignment1.qmd         assignment4_rubric.md   assignment9_rubric.html template1.html
assignment1_rubric.html assignment5.Rnw         assignment9_rubric.md   template1.pdf
assignment1_rubric.md   assignment5.pdf         general_info.tex        template1.qmd
assignment2.Rnw         assignment5_rubric.html index.html              template2.html
assignment2.html        assignment5_rubric.md   index.qmd               template2.pdf
assignment2.pdf         assignment6.Rnw         makefile                template2.qmd
assignment2.qmd         assignment6.pdf         search.json             template3.html
assignment2_rubric.html assignment6_rubric.html simple_template1.html   template3.pdf
assignment2_rubric.md   assignment6_rubric.md   simple_template1.pdf    template3.qmd
assignment3.Rnw         assignment7.Rnw         simple_template1.qmd    template3_files
assignment3.html        assignment7.pdf         simple_template2.html   template4.html
assignment3.pdf         assignment7_fig1.pdf    simple_template2.pdf    template4.pdf
assignment3.qmd         assignment7_rubric.html simple_template2.qmd    template4.qmd
assignment3_rubric.html assignment7_rubric.md   simple_template3.html   tests
assignment3_rubric.md   assignment8.Rnw         simple_template3.pdf
[assignments] quarto render assignment4.qmd                                           gh-pages
ERROR: Include directive failed.
  in file assignment4.qmd,
  could not find file /Users/tedgro/Code/BDA_course_Aalto/assignments/general_info.md.
[assignments]                   

Update the github.io course page to mention and give instructions on the possibility to submit all assignments at the end of the course for a second round of peergrading.

The dmvnorm function checks whether the first argument is a matrix or not. This check is done by checking if class(x) is ‘matrix’. However, class(x) may return a list of class strings. In this case, R decides to do the comparison separately for each element of the list. This will cause an error if the result of the comparison varies between elements.

This bug can be fixed by replacing if (class(x) == ‘matrix’) with if (‘matrix’ %in% class(x)) in the dmvnorm function.

I have attached a small R script that demonstrates this problem. It is a .txt file since GitHub does not support .r files.
bug_demo.txt

We should add sub-sections in each question particularly question 2. For example:

Refactor the following to

2. Include in the report the proposal distribution, number of chains used, the starting points (or the mechanism for generating them), the number of draws generated from each chain, and the warm-up length. Also plot the chains separately for alpha and beta, overlapping chains help to visually assess whether the chains have converged or not. However, in complex scenarios, visual assessment is not sufficient and Rˆ is a more robust indicator of convergence (of the metropolis chains).

2. The following questions pertain to the specifics of the metropolis algorithm:
   a. Was the proposal distribution used, mentioned in the report ?
   b. Were the number of metropolis chains mentioned ?
   ... etc.

This is much easier to refer back to in the rubric.

Task 2: Update the rubric to refer to these individual sub-questions.

There are cases where our test suite for markmyassignment checks the functions with some parameter names that must be present in the definition of the student, as in

p_identical_twin <- function(fraternal_prob=..., identical_prob=...) {}

If the student then names the function arguments using a different name R will raise an error, so we should just state cleary in the assignment pdf that they must use those same names.

Dear Vehtari

1. The script exercises/additional_files/psrf.py is not compatible with Python3. This is because it uses the function xrange, which was removed in Python3. In order to make it Python3-compatible, you can simply have the following block in the top of the script:

try:
    xrange
except NameError:
    xrange = range

However, since the end-of-life Python2 is drawing near, most cental Python projects have vowed to stop supporting Python 2 before 2020, including Numpy, PyStan, Scipy, Jupyter and Cython, and Python3 was released more than 10 years ago, I'd argue it instead might be worth just dropping Python 2 compatibility instead.

2. When given return_extra=True, the output tau is not documented.

Are you supposed to explain the terms with one sentence, or with one sentence each?

We can also simplify a bit the text of the exercise 1.3. It currently stands as

The researchers are happy with these preliminary results (about 97% success rate), and wish to get the test to market as soon as possible. How would you advise them? Base your answer on elementary probability calculus.

We could leave it

The researchers are happy with these preliminary results, and wish to get the test to market as soon as possible. How would you advise them? Base your answer on bayes' rule computations.

Some fo the students where puzzled by the "basic calculus" bit.

ping @MansMeg @kunalghosh

The recommended tool in this course is R (with the IDE R-Studio). You candownload Rhereand R-Studiohere. There are tons of tutorials, videos andintroductions to R and R-Studio online. You can find some initial hints here.

Hi, the link (https://rstudio.com/online-learning/) in You can find some initial hints here. in assignment 1 is outdated.

Hi, I might found a typo in the BDA3 book. Page 44 where there is derivation about the marginal distribution of Poisson data. The final line which shows the prior predictive distribution reduces to a negative binomial density. The one in the book is

Maybe the correct one is the following?

1. The rubric 5 does not have clear distinctions between the exercises, as a result some students decided to abandon the exercise-per-section structure of their reports. -- see that the rest of the rubrics explicitly refer to the exercises in each question, the rubric 5 does not do this;
2. Some things in the rubric were not stated explicitly in the ex5.pdf -- the convergence discussion of the alpha and beta plots.
3. (Suggestion) I would make ex1 and ex2 more distinct. For example, it would be more intuitive if in the ex1 you only need to implement two functions, and in the ex2 you need to do the sampling reporting all relevant parameters (and the plots). Right now, the ex1 needs to do the implementation of two functions and the sampling, but you report the parameters of the sampling only in the ex2, which is not the most straightforward way of doing the experiments: first you do the experiment, and then you define the parameters.

I believe there is a typo on page 10 of Chapter 5 slides,

It seems that install_github fails for some reason when installing the aaltobda package. @andrjohns suggested to use install.packages("aaltobda", repos = c("https://avehtari.github.io/BDA_course_Aalto/", getOption("repos"))) which seems to work.

The error is actually shown in the template, so we should probably change the instructions as it is likely to fail for others too.

From the template:

"
The following installs the aaltobda package:

# Caching should be fine here
install.packages("remotes")

Installing package into '/usr/local/lib/R/site-library'
(as 'lib' is unspecified)
remotes::install_github("avehtari/BDA_course_Aalto", subdir = "rpackage", upgrade="never")

Downloading GitHub repo avehtari/BDA_course_Aalto@HEAD
Error in utils::download.file(url, path, method = method, quiet = quiet,  : 
  download from 'https://api.github.com/repos/avehtari/BDA_course_Aalto/tarball/HEAD' failed

"

https://avehtari.github.io/BDA_R_demos/demos_ch10/demo10_2.html

comment line of scale_color_discrete show red line is g not q

distr <- ggplot(data = df1) +
geom_line(aes(x, p, fill = grp, color = grp)) +
labs(title = 'Target and proposal distributions', x = '', y = '') +
scale_color_discrete(labels = c('q(theta|y)', 'g(theta)')) +
theme(legend.position = 'bottom', legend.title = element_blank())

Dear Aki,

Thank you very much for this great course.

I just wanted to ask if it is possible to render the BDA book in PDF with 1.5 times bigger fonts, or to make epub/mobi/fb2 format?

I just want to make so that the book will display on Kindle DX (10 inch screen reader) with bigger fonts.

Now the fonts are too small, reading on ebook is not comfortable. It is possible only on computer monitor, which is also hard to read for long time (eyes pain).

Thanks in advance.

Regards,
Ihor

The rubrics of exercise 7 problem 2 has very vague definitions of what the model plot for the pooled, separate and hierarchical model should be.
It just mentions

The following model solution plot is used as a reference in the grading for the separate model:

The actual plots need not be included in the rubrics, but maybe a description of what plots are expected would help when we form a report. Something like

The following histograms of a) posterior distribution of mean for 6th machine b) posterior predictive distribution for the 6th machine and c) Posterior distribution of mean for a 7th machine will be used as a reference for the pooled model.

This would then give the student information on what all graphs may be included in the report so that the evaluator will be able to grade better .

Similar things for

discussion about question i)

This could also be written as

discussion about question i) the posterior distribution of mean of measurement of the 6th machine

It seems that the students have some trouble identifying what these mean. Instead of the current explanation we could add "fraternal twin (two different embryos and then could be of different sex) and identical twin (same egg divides into two, so they share the same sex)", or something of the like.

ping @MansMeg @kunalghosh @adhaka

Cross check that the gh-page, peergrade and TA-sheet have the correct deadlines.

Hi,

I am trying to answer some of the questions on the registration form and thus browsing the webpage. I think the links to the schedule in this section

Weekly assignments

There are 9 weekly assignments. Assignments are linked from the schedule and can also be found from git repo assignments folder. The deadline days for the assignments are given in the course schedule.

return 404. (link: https://avehtari.github.io/BDA_course_Aalto/gsu.html#Schedule_2021)

In the description of assignment 1 question 5, it is written "see Exercise 1.5 in BDA3", however it should be Exercise 1.6, please see Bayesian Data Analysis, third edition, page 28 or DBA3

During the exercise sessions the students were confused (at least I had to explain to 4 students) about how to use markmyassignment to validate their solutions.

A note like the following, in exercise PDFs, should be helpful.

Calling mark_my_assignment() without any arguments checks the functions that you are supposed to implement in this assignment.