9. Posterior Distributions for AR(1) Parameters#
We’ll begin with some Python imports.
!pip install arviz pymc numpyro jax
Show code cell output
Requirement already satisfied: arviz in /opt/conda/envs/quantecon/lib/python3.11/site-packages (0.16.1)
Collecting pymc
Obtaining dependency information for pymc from https://files.pythonhosted.org/packages/4a/1b/722890fb21f9910c5a88e5bc716dbd184127e1e36f20b03021a2839655a2/pymc-5.10.3-py3-none-any.whl.metadata
Downloading pymc-5.10.3-py3-none-any.whl.metadata (10 kB)
Collecting numpyro
Obtaining dependency information for numpyro from https://files.pythonhosted.org/packages/f9/e1/545f037aa1d119d83bebb6816d9e01138bb2a2ee5878cdd1c1e76fe3a239/numpyro-0.13.2-py3-none-any.whl.metadata
Downloading numpyro-0.13.2-py3-none-any.whl.metadata (36 kB)
Requirement already satisfied: jax in /opt/conda/envs/quantecon/lib/python3.11/site-packages (0.4.23)
Requirement already satisfied: setuptools>=60.0.0 in /opt/conda/envs/quantecon/lib/python3.11/site-packages (from arviz) (68.0.0)
Requirement already satisfied: matplotlib>=3.2 in /opt/conda/envs/quantecon/lib/python3.11/site-packages (from arviz) (3.7.2)
Requirement already satisfied: numpy<2.0,>=1.21.0 in /opt/conda/envs/quantecon/lib/python3.11/site-packages (from arviz) (1.24.3)
Requirement already satisfied: scipy>=1.8.0 in /opt/conda/envs/quantecon/lib/python3.11/site-packages (from arviz) (1.11.1)
Requirement already satisfied: packaging in /opt/conda/envs/quantecon/lib/python3.11/site-packages (from arviz) (23.1)
Requirement already satisfied: pandas>=1.3.0 in /opt/conda/envs/quantecon/lib/python3.11/site-packages (from arviz) (2.0.3)
Requirement already satisfied: xarray>=0.21.0 in /opt/conda/envs/quantecon/lib/python3.11/site-packages (from arviz) (2023.6.0)
Requirement already satisfied: h5netcdf>=1.0.2 in /opt/conda/envs/quantecon/lib/python3.11/site-packages (from arviz) (1.3.0)
Requirement already satisfied: typing-extensions>=4.1.0 in /opt/conda/envs/quantecon/lib/python3.11/site-packages (from arviz) (4.7.1)
Requirement already satisfied: xarray-einstats>=0.3 in /opt/conda/envs/quantecon/lib/python3.11/site-packages (from arviz) (0.6.0)
Collecting cachetools>=4.2.1 (from pymc)
Obtaining dependency information for cachetools>=4.2.1 from https://files.pythonhosted.org/packages/a2/91/2d843adb9fbd911e0da45fbf6f18ca89d07a087c3daa23e955584f90ebf4/cachetools-5.3.2-py3-none-any.whl.metadata
Downloading cachetools-5.3.2-py3-none-any.whl.metadata (5.2 kB)
Requirement already satisfied: cloudpickle in /opt/conda/envs/quantecon/lib/python3.11/site-packages (from pymc) (2.2.1)
Collecting fastprogress>=0.2.0 (from pymc)
Downloading fastprogress-1.0.3-py3-none-any.whl (12 kB)
Collecting pytensor<2.19,>=2.18.1 (from pymc)
Obtaining dependency information for pytensor<2.19,>=2.18.1 from https://files.pythonhosted.org/packages/58/d4/4431a529ac09aa900df44cf801eaf043f3724e03299f173f44edbff77490/pytensor-2.18.6-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl.metadata
Downloading pytensor-2.18.6-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl.metadata (9.7 kB)
Requirement already satisfied: jaxlib>=0.4.14 in /opt/conda/envs/quantecon/lib/python3.11/site-packages (from numpyro) (0.4.23+cuda12.cudnn89)
Requirement already satisfied: multipledispatch in /opt/conda/envs/quantecon/lib/python3.11/site-packages (from numpyro) (0.6.0)
Requirement already satisfied: tqdm in /opt/conda/envs/quantecon/lib/python3.11/site-packages (from numpyro) (4.65.0)
Requirement already satisfied: ml-dtypes>=0.2.0 in /opt/conda/envs/quantecon/lib/python3.11/site-packages (from jax) (0.3.1)
Requirement already satisfied: opt-einsum in /opt/conda/envs/quantecon/lib/python3.11/site-packages (from jax) (3.3.0)
Requirement already satisfied: h5py in /opt/conda/envs/quantecon/lib/python3.11/site-packages (from h5netcdf>=1.0.2->arviz) (3.9.0)
Requirement already satisfied: contourpy>=1.0.1 in /opt/conda/envs/quantecon/lib/python3.11/site-packages (from matplotlib>=3.2->arviz) (1.0.5)
Requirement already satisfied: cycler>=0.10 in /opt/conda/envs/quantecon/lib/python3.11/site-packages (from matplotlib>=3.2->arviz) (0.11.0)
Requirement already satisfied: fonttools>=4.22.0 in /opt/conda/envs/quantecon/lib/python3.11/site-packages (from matplotlib>=3.2->arviz) (4.25.0)
Requirement already satisfied: kiwisolver>=1.0.1 in /opt/conda/envs/quantecon/lib/python3.11/site-packages (from matplotlib>=3.2->arviz) (1.4.4)
Requirement already satisfied: pillow>=6.2.0 in /opt/conda/envs/quantecon/lib/python3.11/site-packages (from matplotlib>=3.2->arviz) (9.4.0)
Requirement already satisfied: pyparsing<3.1,>=2.3.1 in /opt/conda/envs/quantecon/lib/python3.11/site-packages (from matplotlib>=3.2->arviz) (3.0.9)
Requirement already satisfied: python-dateutil>=2.7 in /opt/conda/envs/quantecon/lib/python3.11/site-packages (from matplotlib>=3.2->arviz) (2.8.2)
Requirement already satisfied: pytz>=2020.1 in /opt/conda/envs/quantecon/lib/python3.11/site-packages (from pandas>=1.3.0->arviz) (2023.3.post1)
Requirement already satisfied: tzdata>=2022.1 in /opt/conda/envs/quantecon/lib/python3.11/site-packages (from pandas>=1.3.0->arviz) (2023.3)
Requirement already satisfied: filelock in /opt/conda/envs/quantecon/lib/python3.11/site-packages (from pytensor<2.19,>=2.18.1->pymc) (3.9.0)
Collecting etuples (from pytensor<2.19,>=2.18.1->pymc)
Downloading etuples-0.3.9.tar.gz (30 kB)
Preparing metadata (setup.py) ... ?25l-
done
?25hCollecting logical-unification (from pytensor<2.19,>=2.18.1->pymc)
Downloading logical-unification-0.4.6.tar.gz (31 kB)
Preparing metadata (setup.py) ... ?25l-
done
?25hCollecting miniKanren (from pytensor<2.19,>=2.18.1->pymc)
Downloading miniKanren-1.0.3.tar.gz (41 kB)
?25l ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 0.0/41.3 kB ? eta -:--:--
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 41.3/41.3 kB 7.9 MB/s eta 0:00:00
?25h
Preparing metadata (setup.py) ... ?25l-
done
?25hCollecting cons (from pytensor<2.19,>=2.18.1->pymc)
Downloading cons-0.4.6.tar.gz (26 kB)
Preparing metadata (setup.py) ... ?25l-
done
?25hRequirement already satisfied: six in /opt/conda/envs/quantecon/lib/python3.11/site-packages (from multipledispatch->numpyro) (1.16.0)
Requirement already satisfied: toolz in /opt/conda/envs/quantecon/lib/python3.11/site-packages (from logical-unification->pytensor<2.19,>=2.18.1->pymc) (0.12.0)
Downloading pymc-5.10.3-py3-none-any.whl (472 kB)
?25l ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 0.0/472.4 kB ? eta -:--:--
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 472.4/472.4 kB 15.3 MB/s eta 0:00:00
?25h
Downloading numpyro-0.13.2-py3-none-any.whl (312 kB)
?25l ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 0.0/312.7 kB ? eta -:--:--
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 312.7/312.7 kB 35.1 MB/s eta 0:00:00
?25hDownloading cachetools-5.3.2-py3-none-any.whl (9.3 kB)
Downloading pytensor-2.18.6-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (2.0 MB)
?25l ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 0.0/2.0 MB ? eta -:--:--
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 2.0/2.0 MB 59.6 MB/s eta 0:00:00
?25h
Building wheels for collected packages: cons, logical-unification, etuples, miniKanren
Building wheel for cons (setup.py) ... ?25l-
\
done
?25h Created wheel for cons: filename=cons-0.4.6-py3-none-any.whl size=9100 sha256=5005dc2360d9ac1e34e484b2542329ffc49005ce805cbe097729a418b656650e
Stored in directory: /github/home/.cache/pip/wheels/95/8f/45/fe0a5b5e232401da571d514eb545833fbe220993ac8336c94e
Building wheel for logical-unification (setup.py) ... ?25l-
\
done
?25h Created wheel for logical-unification: filename=logical_unification-0.4.6-py3-none-any.whl size=13911 sha256=fc1ebb9249420af61e37d4e605cdf9b6251f01c5806cba0bc8d51a504dc0d80c
Stored in directory: /github/home/.cache/pip/wheels/b8/34/a9/c11a21ef1f1b6d2e5ae518dd5d28c0bd2b131c5d6e5d4417c3
Building wheel for etuples (setup.py) ... ?25l-
\
done
?25h Created wheel for etuples: filename=etuples-0.3.9-py3-none-any.whl size=12617 sha256=c4872a9853b84aceac52eae5585387d2273deae9141ad676e54d91d64e1fe71b
Stored in directory: /github/home/.cache/pip/wheels/11/5a/ad/ebef6395e95b1dfd742afb44db6d92bd5e8921a572b9ff8618
Building wheel for miniKanren (setup.py) ... ?25l-
\
done
?25h Created wheel for miniKanren: filename=miniKanren-1.0.3-py3-none-any.whl size=23910 sha256=c4dc051049d2b18eb8611690bef81ef5774403cd8f062d0c635a4862551e78b7
Stored in directory: /github/home/.cache/pip/wheels/8d/86/58/7da4a186a7bb721885e4fb811a54b5a0065b0c624a5deb1259
Successfully built cons logical-unification etuples miniKanren
Installing collected packages: fastprogress, cachetools, logical-unification, numpyro, cons, etuples, miniKanren, pytensor, pymc
Successfully installed cachetools-5.3.2 cons-0.4.6 etuples-0.3.9 fastprogress-1.0.3 logical-unification-0.4.6 miniKanren-1.0.3 numpyro-0.13.2 pymc-5.10.3 pytensor-2.18.6
WARNING: Running pip as the 'root' user can result in broken permissions and conflicting behaviour with the system package manager. It is recommended to use a virtual environment instead: https://pip.pypa.io/warnings/venv
import arviz as az
import pymc as pmc
import numpyro
from numpyro import distributions as dist
import numpy as np
import jax.numpy as jnp
from jax import random
import matplotlib.pyplot as plt
import logging
logging.basicConfig()
logger = logging.getLogger('pymc')
logger.setLevel(logging.CRITICAL)
This lecture uses Bayesian methods offered by pymc and numpyro to make statistical inferences about two parameters of a univariate first-order autoregression.
The model is a good laboratory for illustrating consequences of alternative ways of modeling the distribution of the initial \(y_0\):
As a fixed number
As a random variable drawn from the stationary distribution of the \(\{y_t\}\) stochastic process
The first component of the statistical model is
where the scalars \(\rho\) and \(\sigma_x\) satisfy \(|\rho| < 1\) and \(\sigma_x > 0\); \(\{\epsilon_{t+1}\}\) is a sequence of i.i.d. normal random variables with mean \(0\) and variance \(1\).
The second component of the statistical model is
Consider a sample \(\{y_t\}_{t=0}^T\) governed by this statistical model.
The model implies that the likelihood function of \(\{y_t\}_{t=0}^T\) can be factored:
where we use \(f\) to denote a generic probability density.
The statistical model (9.1)-(9.2) implies
We want to study how inferences about the unknown parameters \((\rho, \sigma_x)\) depend on what is assumed about the parameters \(\mu_0, \sigma_0\) of the distribution of \(y_0\).
Below, we study two widely used alternative assumptions:
\((\mu_0,\sigma_0) = (y_0, 0)\) which means that \(y_0\) is drawn from the distribution \({\mathcal N}(y_0, 0)\); in effect, we are conditioning on an observed initial value.
\(\mu_0,\sigma_0\) are functions of \(\rho, \sigma_x\) because \(y_0\) is drawn from the stationary distribution implied by \(\rho, \sigma_x\).
Note: We do not treat a third possible case in which \(\mu_0,\sigma_0\) are free parameters to be estimated.
Unknown parameters are \(\rho, \sigma_x\).
We have independent prior probability distributions for \(\rho, \sigma_x\) and want to compute a posterior probability distribution after observing a sample \(\{y_{t}\}_{t=0}^T\).
The notebook uses pymc4 and numpyro to compute a posterior distribution of \(\rho, \sigma_x\). We will use NUTS samplers to generate samples from the posterior in a chain. Both of these libraries support NUTS samplers.
NUTS is a form of Monte Carlo Markov Chain (MCMC) algorithm that bypasses random walk behaviour and allows for convergence to a target distribution more quickly. This not only has the advantage of speed, but allows for complex models to be fitted without having to employ specialised knowledge regarding the theory underlying those fitting methods.
Thus, we explore consequences of making these alternative assumptions about the distribution of \(y_0\):
A first procedure is to condition on whatever value of \(y_0\) is observed. This amounts to assuming that the probability distribution of the random variable \(y_0\) is a Dirac delta function that puts probability one on the observed value of \(y_0\).
A second procedure assumes that \(y_0\) is drawn from the stationary distribution of a process described by (9.1) so that \(y_0 \sim {\cal N} \left(0, {\sigma_x^2\over (1-\rho)^2} \right) \)
When the initial value \(y_0\) is far out in a tail of the stationary distribution, conditioning on an initial value gives a posterior that is more accurate in a sense that we’ll explain.
Basically, when \(y_0\) happens to be in a tail of the stationary distribution and we don’t condition on \(y_0\), the likelihood function for \(\{y_t\}_{t=0}^T\) adjusts the posterior distribution of the parameter pair \(\rho, \sigma_x \) to make the observed value of \(y_0\) more likely than it really is under the stationary distribution, thereby adversely twisting the posterior in short samples.
An example below shows how not conditioning on \(y_0\) adversely shifts the posterior probability distribution of \(\rho\) toward larger values.
We begin by solving a direct problem that simulates an AR(1) process.
How we select the initial value \(y_0\) matters.
If we think \(y_0\) is drawn from the stationary distribution \({\mathcal N}(0, \frac{\sigma_x^{2}}{1-\rho^2})\), then it is a good idea to use this distribution as \(f(y_0)\). Why? Because \(y_0\) contains information about \(\rho, \sigma_x\).
If we suspect that \(y_0\) is far in the tails of the stationary distribution – so that variation in early observations in the sample have a significant transient component – it is better to condition on \(y_0\) by setting \(f(y_0) = 1\).
To illustrate the issue, we’ll begin by choosing an initial \(y_0\) that is far out in a tail of the stationary distribution.
def ar1_simulate(rho, sigma, y0, T):
# Allocate space and draw epsilons
y = np.empty(T)
eps = np.random.normal(0.,sigma,T)
# Initial condition and step forward
y[0] = y0
for t in range(1, T):
y[t] = rho*y[t-1] + eps[t]
return y
sigma = 1.
rho = 0.5
T = 50
np.random.seed(145353452)
y = ar1_simulate(rho, sigma, 10, T)
plt.plot(y)
plt.tight_layout()
Now we shall use Bayes’ law to construct a posterior distribution, conditioning on the initial value of \(y_0\).
(Later we’ll assume that \(y_0\) is drawn from the stationary distribution, but not now.)
First we’ll use pymc4.
9.1. PyMC Implementation#
For a normal distribution in pymc,
\(var = 1/\tau = \sigma^{2}\).
AR1_model = pmc.Model()
with AR1_model:
# Start with priors
rho = pmc.Uniform('rho', lower=-1., upper=1.) # Assume stable rho
sigma = pmc.HalfNormal('sigma', sigma = np.sqrt(10))
# Expected value of y at the next period (rho * y)
yhat = rho * y[:-1]
# Likelihood of the actual realization
y_like = pmc.Normal('y_obs', mu=yhat, sigma=sigma, observed=y[1:])
pmc.sample by default uses the NUTS samplers to generate samples as shown in the below cell:
with AR1_model:
trace = pmc.sample(50000, tune=10000, return_inferencedata=True)
with AR1_model:
az.plot_trace(trace, figsize=(17,6))
Evidently, the posteriors aren’t centered on the true values of \(.5, 1\) that we used to generate the data.
This is a symptom of the classic Hurwicz bias for first order autoregressive processes (see Leonid Hurwicz [Hur50].)
The Hurwicz bias is worse the smaller is the sample (see [OW69]).
Be that as it may, here is more information about the posterior.
with AR1_model:
summary = az.summary(trace, round_to=4)
summary
| mean | sd | hdi_3% | hdi_97% | mcse_mean | mcse_sd | ess_bulk | ess_tail | r_hat | |
|---|---|---|---|---|---|---|---|---|---|
| rho | 0.5364 | 0.0709 | 0.4035 | 0.6707 | 0.0002 | 0.0001 | 168990.5957 | 122188.2762 | 1.0 |
| sigma | 1.0102 | 0.1064 | 0.8201 | 1.2126 | 0.0003 | 0.0002 | 179997.0195 | 142155.7267 | 1.0 |
Now we shall compute a posterior distribution after seeing the same data but instead assuming that \(y_0\) is drawn from the stationary distribution.
This means that
We alter the code as follows:
AR1_model_y0 = pmc.Model()
with AR1_model_y0:
# Start with priors
rho = pmc.Uniform('rho', lower=-1., upper=1.) # Assume stable rho
sigma = pmc.HalfNormal('sigma', sigma=np.sqrt(10))
# Standard deviation of ergodic y
y_sd = sigma / np.sqrt(1 - rho**2)
# yhat
yhat = rho * y[:-1]
y_data = pmc.Normal('y_obs', mu=yhat, sigma=sigma, observed=y[1:])
y0_data = pmc.Normal('y0_obs', mu=0., sigma=y_sd, observed=y[0])
with AR1_model_y0:
trace_y0 = pmc.sample(50000, tune=10000, return_inferencedata=True)
# Grey vertical lines are the cases of divergence
with AR1_model_y0:
az.plot_trace(trace_y0, figsize=(17,6))
with AR1_model:
summary_y0 = az.summary(trace_y0, round_to=4)
summary_y0
| mean | sd | hdi_3% | hdi_97% | mcse_mean | mcse_sd | ess_bulk | ess_tail | r_hat | |
|---|---|---|---|---|---|---|---|---|---|
| rho | 0.8757 | 0.0812 | 0.7320 | 0.9985 | 0.0002 | 0.0002 | 107159.5599 | 79689.2489 | 1.0 |
| sigma | 1.4052 | 0.1469 | 1.1443 | 1.6868 | 0.0004 | 0.0003 | 142322.9231 | 113205.7043 | 1.0 |
Please note how the posterior for \(\rho\) has shifted to the right relative to when we conditioned on \(y_0\) instead of assuming that \(y_0\) is drawn from the stationary distribution.
Think about why this happens.
Hint
It is connected to how Bayes Law (conditional probability) solves an inverse problem by putting high probability on parameter values that make observations more likely.
We’ll return to this issue after we use numpyro to compute posteriors under our two alternative assumptions about the distribution of \(y_0\).
We’ll now repeat the calculations using numpyro.
9.2. Numpyro Implementation#
def plot_posterior(sample):
"""
Plot trace and histogram
"""
# To np array
rhos = sample['rho']
sigmas = sample['sigma']
rhos, sigmas, = np.array(rhos), np.array(sigmas)
fig, axs = plt.subplots(2, 2, figsize=(17, 6))
# Plot trace
axs[0, 0].plot(rhos) # rho
axs[1, 0].plot(sigmas) # sigma
# Plot posterior
axs[0, 1].hist(rhos, bins=50, density=True, alpha=0.7)
axs[0, 1].set_xlim([0, 1])
axs[1, 1].hist(sigmas, bins=50, density=True, alpha=0.7)
axs[0, 0].set_title("rho")
axs[0, 1].set_title("rho")
axs[1, 0].set_title("sigma")
axs[1, 1].set_title("sigma")
plt.show()
def AR1_model(data):
# set prior
rho = numpyro.sample('rho', dist.Uniform(low=-1., high=1.))
sigma = numpyro.sample('sigma', dist.HalfNormal(scale=np.sqrt(10)))
# Expected value of y at the next period (rho * y)
yhat = rho * data[:-1]
# Likelihood of the actual realization.
y_data = numpyro.sample('y_obs', dist.Normal(loc=yhat, scale=sigma), obs=data[1:])
# Make jnp array
y = jnp.array(y)
# Set NUTS kernal
NUTS_kernel = numpyro.infer.NUTS(AR1_model)
# Run MCMC
mcmc = numpyro.infer.MCMC(NUTS_kernel, num_samples=50000, num_warmup=10000, progress_bar=False)
mcmc.run(rng_key=random.PRNGKey(1), data=y)
plot_posterior(mcmc.get_samples())
mcmc.print_summary()
mean std median 5.0% 95.0% n_eff r_hat
rho 0.54 0.07 0.54 0.42 0.66 34735.88 1.00
sigma 1.01 0.11 1.00 0.83 1.18 33245.42 1.00
Number of divergences: 0
Next, we again compute the posterior under the assumption that \(y_0\) is drawn from the stationary distribution, so that
Here’s the new code to achieve this.
def AR1_model_y0(data):
# Set prior
rho = numpyro.sample('rho', dist.Uniform(low=-1., high=1.))
sigma = numpyro.sample('sigma', dist.HalfNormal(scale=np.sqrt(10)))
# Standard deviation of ergodic y
y_sd = sigma / jnp.sqrt(1 - rho**2)
# Expected value of y at the next period (rho * y)
yhat = rho * data[:-1]
# Likelihood of the actual realization.
y_data = numpyro.sample('y_obs', dist.Normal(loc=yhat, scale=sigma), obs=data[1:])
y0_data = numpyro.sample('y0_obs', dist.Normal(loc=0., scale=y_sd), obs=data[0])
# Make jnp array
y = jnp.array(y)
# Set NUTS kernal
NUTS_kernel = numpyro.infer.NUTS(AR1_model_y0)
# Run MCMC
mcmc2 = numpyro.infer.MCMC(NUTS_kernel, num_samples=50000, num_warmup=10000, progress_bar=False)
mcmc2.run(rng_key=random.PRNGKey(1), data=y)
plot_posterior(mcmc2.get_samples())
mcmc2.print_summary()
mean std median 5.0% 95.0% n_eff r_hat
rho 0.88 0.08 0.89 0.76 1.00 29063.44 1.00
sigma 1.40 0.15 1.39 1.17 1.64 25674.09 1.00
Number of divergences: 0
Look what happened to the posterior!
It has moved far from the true values of the parameters used to generate the data because of how Bayes’ Law (i.e., conditional probability)
is telling numpyro to explain what it interprets as “explosive” observations early in the sample.
Bayes’ Law is able to generate a plausible likelihood for the first observation by driving \(\rho \rightarrow 1\) and \(\sigma \uparrow\) in order to raise the variance of the stationary distribution.
Our example illustrates the importance of what you assume about the distribution of initial conditions.