Chapter 19 - Sensitivity by Model Band

[flpgrz]

Hi, I have a question on the Sensitivity by Model Band section of Chapter 19.

Why do we compute the sensitivity for each band like this using a single variable linear regression model?

@curry
def sensitivity(data, y, t):
        # line coeficient for the one variable linear regression 
        return (np.sum((data[t] - data[t].mean())*(data[y] - data[y].mean())) /
                np.sum((data[t] - data[t].mean())**2))

I am struggling to understand why this model is used here, while a few cells above we are computing the sensitivity differently using (y(t+h) - y(t)) / h with OLS with interaction terms. For instance, couldn't why compute the mean of the sensitivity we already have for each band?

Thanks