How I Check My Own Research: Five Simple Reproducibility Tests for African Students, Built and Tried on My Own Work

Many published findings cannot be repeated, and students rarely learn how to catch errors before they print them. This proposal describes five checks that any student with a laptop and free Python can run on their own work before submitting it. The checks ask whether every reference exists, whether reported test scores agree with each other, whether a reported average is even possible, whether the computer program finished its job, and whether the data can tell the model's numbers apart. I built the checks into one short script and ran them on my own earlier drafts. They found two references that do not exist, an F1 score of 0.76 that should have been about 0.34, a simulation that quietly stopped halfway, and model parameters that could not be estimated. The proposal plans a workshop for Nigerian students and a test of whether the checks catch planted errors.