Cocoan lab

Bayes Factor Calculator (Bayesian MANOVA)

This website provides a convinient web-based bayes factor calculator for bayesian MANOVA. This website is designed for a specific type of data -- x, y, z, and group information. We originally made this for fMRI data to test whether two conditions or studies activate same or different peak locations.


What's inside?

  • This is using BRMS R package. BRMS stands for Bayesian Regression Models using 'Stan'.
  • For the accurate calculation of bayes factor, setting the prior correctly is crucial. We used weakly-informative priors recommended by Gelman and Hill, 2007 (Ch 13, 17) and STAN manual [v2.17.0] 9.13 and 9.15.
  • You can see the whole code here

Data Format

  • A comma-separated text file with four columns (x, y, z, and group)
  • Each row is observation
  • You can download the example data here
  • It should look like this :

  x,        y,        z,    group
  4,       10,       34,        a
  4,       16,       44,        a
  4,        4,       18,        a
  4,       14,       38,        a
  0,       32,       -4,        a
  2,        2,       44,        a
  0,       36,       -4,        b
  6,       40,       38,        b
  2,        4,       24,        b
  4,        6,       40,        b
  4,       48,       20,        b
  12,      -2,       26,        b
  -16,     36,       -4,        b
  

Analysis

  • Null hypothesis is that xyz coordinates are not different between two groups
  • The alternative hypothesis is that xyz coordinates are different between two groups
  • This web-based calculator provides both BF01 (Bayes Factor in favor of null hypothesis) and BF10 (Bayes Factor in favor of alternative hypothesis)


Note: It takes long (~2 mins)

It will take much longer or cause an error if you run more than two analyses simultaneously.
If you run the analyses more efficiently, please consider to run the analysis on your local computer using this code.


please report issues here

Who made this?

  • Cocoan lab (PI: Choong-Wan Woo)
  • Accompanied article: YongWook Hong, Yejong Yoo, Jihoon Han, Tor D. Wager, Choong-Wan Woo, 2019, False-positive neuroimaging: Undisclosed flexibility in testing spatial hypotheses allows presenting anything as a replicated finding, biorxiv