Mathematical Magician
E-mail: gangchen@mail.nih.gov
Phone: 301-594-9193
FMRI Statistical Analysis in R
- R and package installation
- Intraclass correlation with REML method: 3dICC_REML: NEW
- Multi-Variate Modeling approach to group analysis: 3dMVM
- Mixed-effects meta (or multilevel) analysis: 3dMEMA
- Linear mixed-effects modeling in group analysis: 3dLME
- Structural vector autoregressive (SVAR) analysis: 1dSVAR
- Vector autorgression (VAR) (aka Granger causality analysis): 1dGC
- Independent component analysis: 3dICA
- Kolmogorov–Smirnov test : 3dKS
Connectivity Analysis
- Seed-based analysis (aka functional connectivity)
- 3dGC: Bivariate autoregression
- Context-dependent correlation (aka psychophysiological interaction analysis)
- Simple correlation (outdated)
- ROI-based (or network) analysis (aka effective connectivity)
- 1dSVAR: Structural vector autoregression (SVAR) (New: just released)
- 1dGC: Vector autorgression (VAR) (aka Granger causality analysis)
- 1dSEM: Structural equation modeling (SEM) or path analysis (2007 HBM poster at Chicago) (OLD: not recommended)
- 1dSEMr: Structural equation modeling (SEM) or path analysis (OLD: not recommended)
Some issues in FMRI analysis
- When and how to center a variable?
- How to run group analysis when the BOLD response is modeled with multiple basis functions at individual level
- Scaling and percent signal change: before or after individual subject analsys?
- How to set up contrast weights for trend analysis
- How to construct matrices for main effects and interactions in individual subject analysis with 3dDeconvolve
- How to check model fitting in individual subject analysis
- Basics of group analysis
======================================================
Statistical Calculators
- AFNI command cdf
- Normal, t, F, Chi-Square, Poisson, Binomial, and Exponential distributions
- Conversion between t-statistic and correlation coefficient
- Different types of effects for sums of squares
Matlab Package for Individual and Group Analyses (no longer supported)
This Matlab package is developed under versions 6.5.1 (R13) and 7.0.4 (R14) with Statistical Toolbox (Versions 4.1 and 5.0.2). It comes with the AFNI package, and should be found under your AFNI source code directory, like .../AFNI/src/matlab/. You can also download it as a standalone software from the following link, but try to avoid multiple copies of the files if possible.
- How to download the Matlab script files (updated August 16, 2005)
- How to run ANOVA
- Group analysis with unbalanced designs
- Acknowledgements
Thanks go to the following people who in their unique ways have helped me develop this Matlab package during its birth labor or during its growth with less bugs: Bob Cox, Donald Burrill, Ziad Saad, Rich Hammett, Alex Clark, Jiong-Jiong Yang, Jed Meltzer, Lukas Pezawas, Anders H. Andersen
======================================================
Miscellaneous
Publications