The gracious people at General Assembly allowed me to give a guest lecture on a topic I am quite partial to: generalized linear models (GLMs).
Much of the data used here come from UCLA's IDRE tutorial modules. I claim no rights to owning them. I make no profit from using this data. The rest of the data is open source. I used the famous iris data, as well as the ggplot2-included diamonds data.
The lecture features intermittent code-a-longs. The python-notebooks directory contains Jupyter notebooks for the accompanying Python code. Feel free to clone this repository down and try the code out yourself!
I believe it is important to be language-agnostic. Since R is my native tongue, I have also replicated all of the Python code in the python-notebooks folder into R for reference. This can be found in the R-notebooks folder. The 03 file in the R-notebooks folder contains an example of cumulative logistic regression - something not easily implemented in Python.
I am still considering whether or not I want to upload the full presentation. I am leaning towards no, but feel free to contact me and ask anyway!
However, here is a list of topics I cover in the presentation:
- Recap OLS
- Recap logistic regression
- Discuss the general form of a GLM
- Discuss some common GLMs and their applications
- A brief interlude on model inference
- GLMs for classification
- Logistic regression for classification
- Multinomial regression for classification
- Cumulative logit regression for classification
- GLM Pot Pourri