“And don’t use any math” was my manager’s instruction.
How else am I presupposed to explain how regression works!?
Little did I do know that a giant a part of a knowledge scientist’s job is to elucidate models in non-technical terms. All we want to know a linear model is a regression summary. But this table of numbers will quickly bore (or frighten) a unique audience. To get your point across, you’ll need to make use of more digestible visualisations. So, we’ll explore 8 plots that can make it easier to make linear regression more accessible:
- Residual plots
- Correlation heatmaps
- Weight plots
- Effect plot
- Mean effect plot
- Individual effect plot
- Trend effect plot
- SHAP values for linear models
We’ll see that plots 4 to 7 are particularly useful as they assist explain a model by way of its contribution to a prediction. We “reframe” the outcomes by way of a quantity that everybody understands. For instance, we will go from explaining regression parameters to…