I wanted to use STATA for a kalman filter simulation and wanted to go about it the following way

1. Generate iid error terms 

$$ {e_t}, {n_t}, {u_t} $$

2. Generate three AR1 process like so

$$ \beta_t = a*b_{t-1} + u_t $$ $$ x_t = c_0 + c_1*x_{t-1} + n_t $$ $$ y_t = b_t*x_t + e_t $$

 3. use Kalman filter to predict b_hat

I have been able to use rnorm() to generate the iid error terms, but I am stuck at the second step. I know how to run a state space model to get the kalman filter estimates but AR1 has really tripped me up.


1 Answer 1


Since you are stuck at just estimating AR(1) models I can show you some example code for estimating the equations you gave. Keep in mind that I did not set slopes: $a$ and $c_1$ in the random walk and random walk with drift. Furthermore, the example code is in Python, however it should be self-explanatory and reproducible in STATA.

import numpy as np
import matplotlib.pyplot as plt

# Setting sample size
sample_size = 100

# Vectors with variables of interest
b = np.zeros(sample_size)
x = np.zeros(sample_size)
y = np.zeros(sample_size)

# IID random errors (normally distributed)
u = np.random.normal(size = sample_size)
n = np.random.normal(size = sample_size)
e = np.random.normal(size = sample_size)

# Drift
c_0 = 2

# Simulating each series
for t in range(sample_size):
    b[t] = b[t-1] + u[t]
    x[t] = c_0 + x[t-1] + n[t]
    y[t] = b[t] * x[t] + e[t]

# Plotting the results    
plt.rcParams["figure.figsize"] = [10,10]

plt.title('Random walk: $b_t = b_{t-1} + u_t$')

plt.title('Random walk with drifft: $x_t = c_0 + x_{t-1} + n_t$')

plt.title('$y_t = b_t * x_t + e_t$')


AR(1) process simulation


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.