Common questions about the topics discussed on this site.

What are "Voodoo Correlations" in simple terms?

It's a catchy name for a statistical problem called "non-independent analysis." It refers to finding very high (puzzlingly high) correlations in fMRI studies by first selecting only the brain regions that already show a strong correlation, and then testing them. This method artificially inflates the result.

Does this mean fMRI is "bad science" or "dead"?

Not at all. fMRI is a powerful and valuable tool for understanding the brain. The debate was not about the tool, but about *how* the data was analyzed. The controversy led to much stronger, more reliable statistical methods in the field. It helped the science get better.

What's the difference between p-hacking and scientific fraud?

Fraud is fabricating (making up) data. P-hacking is a "questionable research practice" where researchers analyze their real data in many different ways until they find a "significant" p-value, and then only report the analysis that "worked." It's misleading, but not the same as making up data from scratch.

What is the single best way to improve the reliability of my own research?

Pre-registration. This means publicly documenting your hypothesis, sample size, and analysis plan *before* you collect or look at your data. This prevents p-hacking and HARKing (Hypothesizing After the Results are Known) and is a cornerstone of the Open Science movement.

Is the "Replication Crisis" only happening in psychology?

No. While psychology (and social neuroscience) was one of the first fields to publicly grapple with it, similar issues with replication, low statistical power, and publication bias have been identified in medicine, economics, biology, and many other scientific fields.