Chapter 14

Author: Keith J. Worsley (Montreal Neurological Institute)

Section 14.1. Introduction - Why do we need statistics?
Section 14.2. Modeling the response to the stimulus
Section 14.3. Modeling the random error
14.3.1: Modeling the temporal correlation

Section 14.4. Estimating the response magnitudes
14.4.1: Notation
14.4.2: The fully efficient estimator
14.4.3: The more robust estimator of SPM'99
14.4.4: Estimation in Fourier space
14.4.5: Comparison of the methods

Section 14.5. Estimating the variance
Section 14.6. Detecting an effect
14.6.1: T-tests
14.6.2: F-tests for several contrasts
14.6.3: When to use F-tests

Section 14.7. Setting up the model - an example
14.7.1: A linear intensity effect
14.7.2: A quadratic intensity effect
14.7.3: Intensity as a factor
14.7.4: The design
14.7.5: The baseline or rest condition

Section 14.8. Optimal experimental design
Section 14.9. Estimating the correlation structure
Section 14.10. Spatial smoothing
14.10.1: Scale space
14.10.2: Spatial information

Section 14.11. Estimating the hemodynamic response function
14.11.1: Over-specifying the hemodynamic response function
14.11.2: Misspecifying the hemodynamic response function
14.11.3: Non-linear hemodynamic response and stimulus non-additivity

Section 14.12. Detecting an effect at an unknown location
14.12.1: The maximum test statistic
14.12.2: The maximum spatial extent of the test statistic
14.12.3: Searching in small regions

Section 14.13. Multiple runs, sessions, and subjects
14.13.1: Conjunctions
Section 14.14. Conclusion