psytest documentation

psytest documentation#

This is the documentation for the psytest package, a Python testing framework that applies the methodology of [1] (PSY) to test for the presence of multiple bubbles in a dataset.

A bubble is defined as a period of explosive growth above a unit root in the time series for a consistent period of time.

The main test in the package is the Backward Sup Augmented Dickey Fuller (BSADF) test of PSY. This test consists of applying the Augmented Dickey Fuller (ADF) test to an backward expanding window of the time series at each point in time to identify the presence of bubbles. A bubble is detected once this test rises above the critical value and ends when it returns to the region of non-rejection of the null hypothesis.

The benefits of this test relative to other tests in the literature is its ability to identify multiple bubbles in a dataset.

The main class of the package is the psytest.PSYBubbles class, which contains methods to calculate the BSADF test statistics and critical values as well as finding the start and end dates of the bubbles.

Installation#

You can install the package from my GitHub repository using pip:

pip install git+https://github.com/joseparreiras/psytest

or clone the repository and install it locally:

git clone https://github.com/joseparreiras/psytest
cd psytest
pip install .

Usage Example#

The package is designed to be easy to use. The main class is psytest.PSYBubbles, which contains methods to calculate the BSADF test statistics and critical values as well as finding the start and end dates of the bubbles.

Assuming the time series data is stored in a numpy.ndarray, you can innitiate the class with the data and the desired parameters:

import numpy as np
from psytest import PSYBubbles

# Create an instance of the PSYBubbles class
psy = PSYBubbles(
   data,
   minwindow=10,
   minlength=90,
   maxlag=2,
)

# Find the start and end dates of the bubbles
psy.find_bubbles(alpha=0.05)

See Replication Exercise for a replication example of the original paper as well as a walkthrough of the package. Also see the psytest for a detailed description of the methods and attributes of the psytest.PSYBubbles class.

Table of Contents

References#

[1]

Peter CB Phillips, Shuping Shi, and Jun Yu. Testing for multiple bubbles: historical episodes of exuberance and collapse in the s&p 500. International economic review, 56(4):1043–1078, 2015.

[2]

Peter CB Phillips, Shuping Shi, and Jun Yu. Testing for multiple bubbles: limit theory of real-time detectors. International Economic Review, 56(4):1079–1134, 2015.

Contents