Statistical inference is concerned with the question of quantifying statistical uncertainty of our estimators of interest, that is, how much can we learn about the parameter of interest from a data sample of a finite length. It tells us how much should/could we trust the results that we get.

The first part is related to Maximum likelihood estimation. An approach to estimation that is based on a likelihood principle, the parameter values are chosen to be most compatible with the observed data sample. Maximum likelihood estimator has some very favourable properties.

In the second part we talk about the computational class of methods for statistical inference based on resampling, the Bootstrap. This is a very general principle that can be applied in variety of settings when other approaches to statistical inference are too complex or not feasible.

The R file with examples can be downloaded here.

In the last part of the series of lectures in the second day, a short introduction to another topic of this course - how to recover causal relationship from observational data, is to be presented. In other words how to tell correlation from causation. Careful distinction between the two is one of the main goals of econometrics. Economists are well trained for this task. The presentation will review some of the popular and successful examples where we could recover causal relationships even without an experiment.

This might give you an idea what will be covered later. We will look at the econometric techniques that are suitable (not only) for these purposes.

Assignment 1 deadline is 4.11.2022

Or you can dowload it here: 

Please make use of the following tool to submit your work.

Here is the video with the solutions to the first assignment:

and the code