Ensuring Reproducibility in Statistical Analysis with Seeds

Never share your findings without ensuring reproducibility.

Always use a ‘seed’ 🌱

To ensure reproducibility in statistical analysis, consider leveraging numpy and scipy. If you’re using scipy.stats to define random variables for an uncertainty analysis study, be sure to specify a Random Number Generator (RNG) object when calling the .rvs() method to generate samples from your distribution.

In the example below, I demonstrate creating an RNG object with numpy using a predefined seed number. Then, I import a normal random variable from scipy.stats, define its parameters and include the random_state argument, which references the previously created RNG.

By establishing a specific starting point (seed) for random number generation, you empower others to replicate the exact conditions of your study. This transparency fosters trust and reliability in scientific discoveries.