P-Hacking: A Practical Guide @Neuro_Skeptic neuroskeptic@googlemail.com http://blogs.discovermagazine.com/neuroskeptic
The Nine Circles of Dante’s Inferno First Circle: Limbo Second Circle: Lust Third Circle: Gluttony Fourth Circle: Greed Fifth Circle: Anger Sixth Circle: Heresy Seventh Circle: Violence Eighth Circle: Fraud Ninth Circle: Treachery
The Nine Circles of Scientific Hell First Circle: Limbo Second Circle: Overselling Third Circle: Post-Hoc Storytelling Fourth Circle: P-Value Fishing Fifth Circle: Creative Use of Outliers Sixth Circle: Plagiarism Seventh Circle: Non-Publication of Data Eighth Circle: Partial Publication of Data Ninth Circle: Inventing Data
P-Fishing Fourth Circle: P-Value Fishing “Those who tried every statistical test in the book until they got a p value less than 0.05 find themselves here, an enormous lake of murky water. Sinners sit on boats and must fish for their food. Fortunately, they have a huge selection of different fishing rods and nets (brand-names include Bayes, Student, Spearman and many more). Unfortunately, only one in 20 fish are edible, so they are constantly hungry.”
P-Fishing Also known as… Questionable Research Practices (QRPs)) Torturing the data P-Hacking Outcome reporting bias Undisclosed flexibility Researcher Degrees of Freedom …and more.
P-Hacking Works! Collect some data
P-Hacking Works! Collect some data Try many statistical tests on the same data
P-Hacking Works! Collect some data Try many statistical tests on the same data Or try many variants of the same data (e.g. removing ‘outliers’.)
P-Hacking Works! Collect some data Try many statistical tests on the same data Or try many variants of the same data (e.g. removing ‘outliers’.) Or try looking at different variables within the dataset
P-Hacking Works! Collect some data Try many statistical tests on the same data Or try many variants of the same data (e.g. removing ‘outliers’.) Or try looking at different variables within the dataset Report the analyses that give the most favourable results (usually the lowest p- values).
Why Hacking Is So Useful There are many choices (‘researcher degrees of freedom’) in data analysis. For example, in a simple task-based fMRI data analysis, Joshua Carp found 7000 combinations of parameters (very conservative). Carp, J. (2012).On the plurality of (methodological) worlds: estimating the analytic flexibility of fMRI experiments Frontiers in Neuroscience
How Not To Get Caught There are many tools, but perhaps the most intuitive: the p curve Try it now! http://www.p-curve.com/app2/ Simonsohn, U. Nelson, L. D. Simmons, J. P. (2013). P-curve: a key to the file-drawer. Journal of Exp. Psychol General
How It Can Be Stopped Smulders YM (2013). A two-step manuscript submission process can reduce publication bias. Journal of Clinical Epidemiology Chambers CD (2013). Registered Reports: a new publishing initiative at Cortex Cortex
Happy hacking! @Neuro_Skeptic neuroskeptic@googlemail.com http://blogs.discovermagazine.com/neuroskeptic