Prediction (out of sample)

python

%matplotlib inline

python

import numpy as np
import matplotlib.pyplot as plt

import statsmodels.api as sm

plt.rc("figure", figsize=(16, 8))
plt.rc("font", size=14)

Artificial data

python

nsample = 50
sig = 0.25
x1 = np.linspace(0, 20, nsample)
X = np.column_stack((x1, np.sin(x1), (x1 - 5) ** 2))
X = sm.add_constant(X)
beta = [5.0, 0.5, 0.5, -0.02]
y_true = np.dot(X, beta)
y = y_true + sig * np.random.normal(size=nsample)

Estimation

python

olsmod = sm.OLS(y, X)
olsres = olsmod.fit()
print(olsres.summary())

In-sample prediction

python

ypred = olsres.predict(X)
print(ypred)

Create a new sample of explanatory variables Xnew, predict and plot

python

x1n = np.linspace(20.5, 25, 10)
Xnew = np.column_stack((x1n, np.sin(x1n), (x1n - 5) ** 2))
Xnew = sm.add_constant(Xnew)
ynewpred = olsres.predict(Xnew)  # predict out of sample
print(ynewpred)

Plot comparison

python

import matplotlib.pyplot as plt

fig, ax = plt.subplots()
ax.plot(x1, y, "o", label="Data")
ax.plot(x1, y_true, "b-", label="True")
ax.plot(np.hstack((x1, x1n)), np.hstack((ypred, ynewpred)), "r", label="OLS prediction")
ax.legend(loc="best")

Predicting with Formulas

Using formulas can make both estimation and prediction a lot easier

python

from statsmodels.formula.api import ols

data = {"x1": x1, "y": y}

res = ols("y ~ x1 + np.sin(x1) + I((x1-5)**2)", data=data).fit()

We use the I to indicate use of the Identity transform. Ie., we do not want any expansion magic from using **2

python

res.params

Now we only have to pass the single variable and we get the transformed right-hand side variables automatically

python

res.predict(exog=dict(x1=x1n))