6.1 One syntax to rule them all

PyMC has a very simple and expressive syntax that allows us to build arbitrary models. That’s usually a blessing, but it can be a burden too. Bambi instead focuses on regression models, and this restriction leads to a more focused syntax and features, as we will see.

Bambi uses a Wilkinson-formula syntax similar to the one used by many R packages like nlme, lme4, and brms. Let’s assume data is a pandas DataFrame like the one shown in Table 6.1.

	y	x	z	g
0	-0.633494	-0.196436	-0.355148	Group A
1	2.32684	0.0163941	-1.22847	Group B
2	0.999604	0.107602	-0.391528	Group C
3	-0.119111	0.804268	0.967253	Group A
4	2.07504	0.991417	0.590832	Group B
5	-0.412135	0.691132	-2.13044	Group C

Table 6.1: A dummy pandas DataFrame

Using this data, we want to build a linear model that predicts y from x. Using PyMC, we would do something like the model in the following code block:

Code 6.1

with pm.Model() as lm: 
  ...