In Assignments 5 and 6, we ask the students to sample from the same Bioassay posterior with Metropolis-Hastings and Stan, respectively.
In neither of these assignments is reporting eg. the posterior mean required. I think it would make solving the assignment more straightforward, and grading more fair.
Assignment 6
For example, in Assignment 6 we grade according to this:
Does the implemented Stan-model seem to be working?
No implementation
Model implemented but results not visualized/reported (1p)
Model implemented, but the results seem weird (1p)
Model seems to work correctly
It's sometimes difficult to give a fair grade, especially if the results are not numerically summarized (which we don't require).
I think a better and fairer grading would be
Does the implemented Stan-model yield correct results?
No implementation (0p)
Model implemented but results not visualized/reported (1p)
Results are incorrect due to a minor mistake (2p)
Yes, and the posterior means are roughly [...] for alpha and [...] for beta (3p)
Moreover, we would add Report the results of the sampling, including posterior means for alpha and beta to task 1.
Assignment 5
In assignment 5, we could simply add a question
Are the posterior means roughly [...] for alpha and [...] for beta?
No (0p)
Yes (1p)
Moreover, we would add Also report the posterior means for alpha and beta to task 2 or 4.
I think this are two excellent ideas. Then we can just compute the true value in the teacher repo to compare with. I think this might be good if you would do a PR to this repo and to the teacher repo.
In Assignments 5 and 6, we ask the students to sample from the same Bioassay posterior with Metropolis-Hastings and Stan, respectively.
In neither of these assignments is reporting eg. the posterior mean required. I think it would make solving the assignment more straightforward, and grading more fair.
Assignment 6
For example, in Assignment 6 we grade according to this:
It's sometimes difficult to give a fair grade, especially if the results are not numerically summarized (which we don't require).
I think a better and fairer grading would be
Moreover, we would add
Report the results of the sampling, including posterior means for alpha and beta
to task 1.Assignment 5
In assignment 5, we could simply add a question
Moreover, we would add
Also report the posterior means for alpha and beta
to task 2 or 4.