Spm12manual, Statistics section (8-10) Notes

FMRI Model Specifictaion

GLM based

The following steps are included: ① explicit GLM design matrix, ② estimation of GLM parameters using classical or Bayesian methods, ③ using the contrast vector to check the results, generate statistical parameter graphs (SPMS) and a posteriori probability plot (PPMs).

Design Matrix: Each row corresponds to a scan, and each column corresponds to an impact variable or explanatory variable (regression amount or explanatory variable). There are a number of related articles on the GLM (general linear model) network,

For multi-Test analysis, SPM is usually divided into two stages (level):

①1st-level Analysis: Individual analyses (within-subject). How much 1st-level analysis is usually performed. Use the Specify FI rst level and Estimate options. The results of these options can be used to do case study.

②2nd-level Analysis:rfx Analysis (Random effect analysis). The case of the subject of the subjects was inferred from the trial.

1st Level Analysis:

①conditions:

Name (1*n cell): condition name;

Onsets (1*n cell): Each cell is a time vector that holds the condition start time, each column corresponds to a condition, each row corresponds to its occurrence time (supplemental, Heart drops Csddzh: The number of scans for task stimulation initiation (e.g. 1:14:70, Representing the task starting from the 1th tr, each 14 tr is a cycle, a total of 70 tr);

Durations: Event duration.

Generally this information is stored in the Condition.mat file, when used directly load comes in, using multiple Conditions.

②regressors

A column in the design matrix used to make a convolution with the blood dynamic response. After the Leave-one-out experiment to use.

It is not clear whether to use multiple regressors for the time being.

③factorial Design

Setting this option allows the SPM to directly generate the necessary test:f-test between the main effect and interaction (1st level) and the contrast image of the participants (2nd level) and T-test.

According to the Crayon video: The trial according to three kinds of product color choice to buy or not to buy, then there are two factor:color,3;choice,2.

SPM default the first factor changes the slowest, the second factor changes the second slow ... And so on

Factorial design is usually used when there is a complex situation, and it is easy to use without setting.

Other factors do not look very much, because in the crayon video is the default.

=============================================

FMRI model Estimation

Model parameters can be estimated using classical (reml-restricted Maximum likelihood) or Bayesian algorithms.

Crayon Video This step directly uses the previous step to generate (? ) is run by default after the Spm.mat.

Generates a Spm.mat that contains the design and estimated model parameters.

See Result: Main panel result.

=============================================

Factorial design Specification (2nd level analysis)

Configure the design matrix to describe the GLM, data, and necessary parameters (stored in GLM) required for statistical analysis. Used to prepare for 2nd level analysis.

Specify 2nd level analysis.

Spm12manual, Statistics section (8-10) Notes