IITM BS Statistical Computing (BSMA3014): Syllabus and Tips

Statistical Computing (BSMA3014) is a 4 credit degree level elective on how computers do statistics. You learn to generate random samples, estimate with Monte Carlo methods, and fit models with optimisation algorithms such as Newton's method, gradient descent and EM. The course uses R. There are no prerequisites. The instructor is Dootika Vats of IIT Kanpur, listed on the course page as Associate Professor in the Department of Statistics and Data Science.

The course page and the handbook agree. The handbook's degree level table (updated 18 March 2026) lists it as a DEGREE course with the SE tag, 4 credits and no prerequisite.

What you learn

• Weeks 1 to 4: making random samples. An introduction to R and to Monte Carlo, pseudorandom numbers, and ways to sample from a distribution: the inverse transform, accept and reject, composition, Box-Muller, ratio of uniforms, and sampling from distributions with many variables.
• Weeks 5 and 6: importance sampling. Simple and weighted importance sampling, with their bias, variance and consistency, and how to pick a good proposal. Then a review of likelihood functions and maximum likelihood (MLE).
• Weeks 7 to 9: optimisation for estimation. Linear regression as MLE, penalised regression, what to do when the MLE has no closed form, Taylor series, Newton's method, gradient descent for logistic regression, the MM algorithm with bridge regression, and the EM algorithm with Gaussian mixture models.
• Weeks 10 to 12: resampling and stochastic methods. EM for Gaussian mixtures, cross-validation, bootstrapping, stochastic gradient descent and simulated annealing.

Several weeks list "examples and code" or "with code", so expect to write and run R, not just read about methods.

How it is assessed

The course page lists 12 weeks of coursework, weekly online assignments, 2 in-person invigilated quizzes and 1 in-person invigilated end term exam. It does not mention an OPPE or a project.

Where it counts

• Degree level elective. It fills 4 elective credits. See the DS electives list.
• BSc elective credits. Its 3xxx code makes it a level 3 course under the handbook's rule. Level 3 credits can fill the 8 BSc elective credits. See the BSc degree level.
• No minor. It is not in any minor list.
• Term plan. The March 2026 table marks it as offered only in January 2027, out of May 2026, September 2026 and January 2027.

Who finds it hard and how to prepare

Two kinds of students may struggle. Some are fine with theory but slow at turning an algorithm into working R code. Others code well but have forgotten density functions and expectations. The algorithms are short, but a small mistake can give samples that look fine and are wrong.

• Learn basic R. Vectors, functions, loops, and plotting a histogram. That covers most of what the early weeks need.
• Revise Stats 2. Density functions, expectation and variance come up from week 1.
• Revise Maths 2 weeks 9 to 11. Gradients and the Hessian are the heart of Newton's method and gradient descent.
• Always check your sampler. Draw many samples, plot a histogram, and lay the true density over it. If they do not match, find out why before you move on.
• Fix a random seed so you can repeat your results while debugging.
• Use the suggested books. Simulation by Sheldon Ross (fifth edition), and Monte Carlo Statistical Methods by Christian Robert and George Casella.

What to take before and after

• Before: Machine Learning Foundations and Machine Learning Techniques. MLF covers optimisation and ends with EM, and MLT covers Gaussian mixtures with EM and ridge and LASSO regression. This course comes back to the same ideas and makes you code them.
• Related: Linear Statistical Models gives the theory of regression. In the March 2026 table the two run in different terms (September 2026 and January 2027), so you can take them one after the other.

Common questions

Can I use Python instead of R?

The course page introduces R in week 1 and does not mention Python. Assume assignments use R unless your grading document says otherwise.

Is this course more maths or more coding?

Both. The topics are statistical methods, and several weeks pair them with code. If you enjoy seeing an idea work on a computer, this is a good fit.

Why learn sampling methods when libraries already do it?

Libraries hide the method. Knowing how accept and reject or importance sampling works tells you when a result can be trusted, and how to sample from a distribution no library supports.