In the previous chapter, we used a function from NumPy to calculate the parameters.
Now we will use the class object instead of the function to represent the linear regression. This approach takes more lines of code to find the parameters, but it stores a lot of helpful information inside the object and makes the prediction more straightforward.

Building a Linear Regression model

In statsmodels, the OLS class can be used to create a linear regression model.

We first need to initialize an OLS class object using sm.OLS(y, X_tilde). Then train it using the fit() method.

Which is equivalent to:

Note

The constructor of the OLS class expects a specific array X_tilde as an input, which we saw in the Normal Equation. So you need to convert your X array to X_tilde. This is achievable using the sm.add_constant() function.

Finding parameters

When the model is trained, you can easily access the parameters using the params attribute.

Making the predictions

New instances can easily be predicted using predict() method, but you need to preprocess the input for them too:

Getting the summary

As you probably noticed, using the OLS class is not as easy as the polyfit() function. But using OLS has its benefits. While training, it calculates a lot of statistical information. You can access the information using the summary() method.

That's a lot of statistics. We will discuss the table's most important parts in later sections.

Choose the INCORRECT statement.

Select the correct answer

You need to use sm.add_constant() on X before passing it to the OLS() constructor.

You need to use sm.add_constant() on X_new before passing it to the predict().

The regression model is trained using the sm.OLS(y,X_tilde).

Regression's parameters are available using the params attribute.

Submit Answer