Statistics
- P-values and reproducibility issues by JB Poline within Neurohackademy 2018 (1 hr 1 min)
- The evil p value by JB Poline within Neurohackweek 2017 (1 hr 2 min)
- Statistical Decision Theory by Joshua Vogelstein within Brainhack-Vienna (starts at 8 min, ends at 48 min)
- A reminder on how random field theory is used to correct for multiple comparison here.
- A primer on permutation testing (not only) for MVPA by Carsten Allefeld at OHBM 2018 (36 min)
- Cross-validation: what, which and how? by Pradeep Reedy Raamana at OHBM 2018 (30 min)
- Daniel Lakens MOOC on coursera on how to improve your statistical inferences
- Statistical thinking for the 21st century by Russ Poldrack: "I am trained as a psychologist and neuroscientist, not a statistician. However, my research on brain imaging for the last 20 years has required the use of sophisticated statistical and computational tools, and this has required me to teach myself many of the fundamental concepts of statistics. Thus, I think that I have a solid feel for what kinds of statistical methods are important in the scientific trenches."
- Principles, Statistical and Computational Tools for Reproducible Data Science Intermediate-level course by Harvard University in edX
- Gentle Introduction to Bayesian Stats from bayestestR by easystats
A list of R based web based apps from shiny apps and R psychologist to help better understand:
- p-values
- confidence intervals
- p curves and why with a decent power and a large effect size, it is relatively unlikely to find a value between p<.01 and p<.05
- null hypothesis significance testing
- p hacking
- positive predictive values
Partial least squares regression:
This is particularly useful for highdimensional neuroimaging data (particularly when correlating with genetic/transcriptome data). It finds a linear regression model by projecting the predicted variables and the observable variables to a new space. Useful resources if you are new to PLS-R: