Back to all Advice

My 20 published mistakes (and counting)

August 26, 2025

Surprising results are either breakthroughs or bugs. Under pressure, it is easy to declare victory and publish them. Responsible scholars do the opposite: assume they are bugs until you’ve eliminated all other possibilities.

My group works hard to scrutinize results, document carefully, and publish data and code so others can reproduce our work. Most of the time we get this right. Despite this, mistakes sometimes slip through.

I often hear from readers having trouble reproducing one of our results. Our source code resolves most queries, but sometimes they have found a true problem.

Here are issues we have publicly corrected:

• A correction to numerical results, caused by an error in our input data: https://doi.org/10.1177/87552930241245554. Detected by readers who couldn’t independently reproduce our figures.

• An error in reported regression coefficients, caused by a data selection problem: https://doi.org/10.1785/0120190236. Detected by readers who independently implemented our analysis and got slightly different results.

• An error in tabulated results, caused by an algorithm implementation problem: https://doi.org/10.1002/eqe.3233. Detected by readers who were extending our work and noticed a discrepancy.

• A typo in a published equation: https://www.jackwbaker.com/Publications/Shahi_Baker_(2011)_Pulse_PSHA,_BSSA-erratum.pdf. We noticed this internally when extending the work.

• A typo in a published equation: https://www.jackwbaker.com/Publications/Baker_(2007)_EESD_vector_IM_methods_erratum.pdf. We noticed this internally when writing a similar equation in a follow-up paper.

• Fifteen typos in our textbook: https://www.dropbox.com/scl/fi/9md2if9lpdyxsb4flt7gj/Errata.pdf?rlkey=9l54xyxcp0wfw1jiinwvdrcdn&dl=0. Detected by readers who were unsure about our text when studying the book.

These are all fairly minor in that they don’t affect our primary conclusions, though a few did impact numerical results. Others were typographic and had little practical impact, but they were confusing enough that readers reached out.

What have I learned from these errors?

Occasional mistakes are inevitable. Don’t let them happen because of sloppiness or lack of effort. But if a few sneak through, that shouldn't be a source of shame. Publishing corrections hasn’t harmed our reputation. If anything, it has improved it. Readers can see that we take our work seriously, and know we surface problems rather than bury them.

Not all mistakes are equal. Typographical errors are the easiest to commit, as these errors only arise in writing and can’t be detected from accompanying calculations. Readers often identify these discrepancies as inconsistencies between your code and your writing/equations, and they usually don’t cause serious problems. Careful paper editing, including by peers, is needed to prevent these. Data and coding errors are more serious, and also possible because you don’t always have a benchmark correct result to check against. They can be managed by repeating known analyses on the data to confirm reproducibility, testing your code on problems with known solutions, using unit tests, etc. Methodological errors are most serious and are hopefully always avoided through careful research practices and maintaining a skeptical mindset.

It is a compliment that people find problems. I’m glad people are studying our work so carefully that they detect our errors. There are countless non-useful or non-understandable research products in the world that never get used, and their problems will thus never be noticed. If you get a bug report, thank the reporter, fix the issue, and credit them publicly. All of my above corrections include an acknowledgement thanking the person who helped identify it.

Do your best to catch any mistakes before you go to press. Organizing and publishing your data and source code will help you with this. If you’ve been careful and a mistake slips through, publish a correction and don’t worry about it further. Occasional corrections are a sign of progress rather than a symptom of incompetence.