pyMannKendall

Warnings

• gotcha The original Mann-Kendall test assumes serially independent data. Applying it to autocorrelated time series can lead to an inflated Type I error rate (i.e., detecting a trend where none exists due to correlated noise), resulting in misinterpretation of p-values.
  Fix: For autocorrelated data, use modified Mann-Kendall tests available in the package, such as `hamed_rao_modification_test`, `yue_wang_modification_test`, or `pre_whitening_modification_test` that account for serial correlation.
• gotcha The original Mann-Kendall test does not account for seasonality in time series data. Applying it directly to seasonal data can lead to erroneous trend detection or misinterpretation.
  Fix: For data exhibiting seasonal patterns, utilize the `seasonal_test` or `seasonal_sens_slope` functions, which are designed to handle seasonal effects by considering each seasonal period separately.
• gotcha Recent versions (v1.4.2, v1.4.3) have addressed bugs related to type data conversion and integer type data arrays. While fixed, this indicates potential sensitivity to input data types or formats in older versions or edge cases.
  Fix: Ensure your input data is in a suitable numerical format (e.g., NumPy array or Pandas Series of float/int). Always update to the latest version of `pymannkendall` to benefit from bug fixes related to data handling.
• gotcha Reliability of the Mann-Kendall test statistics decreases with very small sample sizes (e.g., 3 or 4 observations).
  Fix: It is recommended to have at least 10 observations for more reliable test results.

Install

• pip install pymannkendall Install with pip

Imports

• pymannkendall

  import pymannkendall as mk

  Standard convention for importing the library, often aliased as 'mk'.
• original_test

  mk.original_test(data)

  Call directly on the aliased 'mk' object after import.

Quickstart

This quickstart demonstrates how to perform a basic Mann-Kendall trend test on a NumPy array. The `original_test` function returns a named tuple with various statistics including the trend ('increasing', 'decreasing', or 'no trend'), p-value, Kendall Tau, and Sen's slope.

import numpy as np
import pymannkendall as mk

# Sample time series data
data = np.array([1, 2, 3, 4, 5, 4, 3, 2, 1, 2, 3, 4, 5])

# Perform the original Mann-Kendall trend test
result = mk.original_test(data)

print(f"Trend: {result.trend}")
print(f"P-value: {result.p}")
print(f"Kendall Tau: {result.Tau}")
print(f"Slope: {result.slope}")
# Output will indicate if a trend is present, its p-value, Tau, and Sen's slope.