where C is an s P matrix, and P = total number <strong>of</strong> parameters in the full model. It canbe shown (see Ref. [2]) that the least squares estimate for which satises the above set<strong>of</strong> linear constraints is given by:b R = b F ; ; X t X ;1 C t C ; X t X ;1 C t ;1CbFwhere b F = least squares estimate <strong>of</strong> under the full model, and b R = least squaresestimate <strong>of</strong> under the reduced model. The error sum <strong>of</strong> squares for this reduced modelis then:SSE(R) =(Z ; Xb R) t (Z ; Xb R)Finally, the test statistic for the general linear test (GLT) is:F =SSE(R) ; SSE(F )df R ; df FSSE(F )df Fwhich has the F (df R ; df F df F ) distribution under the null hypothesis (Ref. [2]), wheredf R ; df F = s.Writing the s P matrix C in terms <strong>of</strong> row vectors:C =264c t 1c t 2.c t s375we can estimate C by:Cb F =2643c t 1b Fc t 2b F7.c t sb F25 643L 1L 27. 5L swhere L 1 , L 2 , :::, L s are the s linear combinations <strong>of</strong> the parameter vector b F speciedby the GLT.Program 3dDeconvolve calculates the s linear combinations L 1 , L 2 , :::, L s , and theGLT F statistic, for each operator specied GLT, and for each voxel, and appends thesevalues as sub-bricks <strong>of</strong> an AFNI \bucket" dataset. (See the Examples in Section 1.4.4.)1.2.12 Concatenation <strong>of</strong> RunsMany users choose to concatenate runs prior to time series analysis. This is a verydelicate operation, due to the implicit time dependence across runs. That is, consecutiveimage volumes are assumed to have been acquired at consecutive points in time but for16
concatenated runs, this is not the case. In order to facilitate analysis <strong>of</strong> concatenated runs,program 3dDeconvolve has been modied to allow theuser to indicate the volume indices,<strong>of</strong> a concatenated dataset, at which the dierentrunsbegin. Internally, the program makesallowance for the fact that an input stimulus from one run should not eect the measuredresponse for the following run.Another change that is required for processing <strong>of</strong> concatenated runs is the addition <strong>of</strong>separate baseline parameters for each run. That is, if the baseline is modeled by a constantplus linear trend, then a separate constant and slope are estimated for each run. Thiswould be necessary even if program 3dTcat had been used to remove the linear trend fromeach run prior to concatenation. The reason is that the stimulus functions are seldomorthogonal to a constant plus linear trend. Hence, removing the linear trends, in isolation,does not yield the same result as the complete least squares solution.As a consequence, if the baseline model is to represent a constant oset plus linear trend(i.e., 2 parameters) for each run, and if the concatenated dataset contains r runs, then thebaseline model contains a total <strong>of</strong> 2r parameters.Suppose that the input dataset is a concatenation <strong>of</strong> 3 runs each run has length N thesingle input stimulus function f is present at time lags 0 through p. Since the individualruns are modeled by:Z n = 10+ 11n + h 0 f n + h 1 f n;1 + h p f n;p + " n , n = p p +1::: N ; 1, for Run #1Z n = 20 + 21 n + h 0 f n + h 1 f n;1 + h p f n;p + " n , n = p p +1::: N ; 1, for Run #2Z n = 30 + 31 n + h 0 f n + h 1 f n;1 + h p f n;p + " n , n = p p +1::: N ; 1, for Run #3the single concatenated run can be modeled by:Z n = ; 10 + 11 n X 1 + ; 20 + 21 (n ; N) X 2 + ; 30 + 31 (n ; 2N) X 3+h 0 f n + h 1 f n;1 + h p f n;p + " n ,n = p p +1::: N ; 1,N + p N + p +1::: 2N ; 1,2N + p 2N + p +1::: 3N ; 1,whereX 1 =X 2 =X 3 = 1 if time point isfromRun #10 otherwise 1 if time point isfromRun #20 otherwise 1 if time point isfromRun #30 otherwiseIn this notation, ij is the coecient <strong>of</strong> t j for the ith run.Therefore, the matrix representation <strong>of</strong> the full model is:Z = X + ":17
- Page 1: Deconvolution Analysis of FMRI Time
- Page 7 and 8: Since the system is linear, the ope
- Page 9: the above equation can be written:Z
- Page 13 and 14: where s(h k ) is the square root of
- Page 15: Here, SSE(F )isthe error sum of squ
- Page 19: This equation relates the output y
- Page 22 and 23: value is pnum = 1 (corresponding to
- Page 24 and 25: model parameters, and P columns, ea
- Page 26 and 27: Program 3dDeconvolve Command Line f
- Page 28 and 29: f(t) = f 0 0 1 1 1 0 1 1 0 0 1 0 1
- Page 30 and 31: If we model the measured response b
- Page 32 and 33: Note that, since there is only one
- Page 34 and 35: Full Model:MSE = 0.9618R^2 = 0.9835
- Page 36 and 37: that are not of inherent interest,
- Page 38 and 39: Program 3dDeconvolve Command Line f
- Page 40 and 41: Now, model the measured data as a c
- Page 42 and 43: -stim label 1 Random -stim label 2
- Page 44 and 45: Note: This example is for illustrat
- Page 46 and 47: The output is the same as for Examp
- Page 48 and 49: After program 3dDeconvolve has nish
- Page 50 and 51: Program 3dDeconvolve Command Line f
- Page 52 and 53: Program:3dDeconvolveAuthor:B. Dougl
- Page 54 and 55: cat y.1D y.1D > ycat.1DBe sure to r
- Page 56 and 57: Recall that in Example 1.4.4.6, we
- Page 58 and 59: and the noisy measurement data:wp(t
- Page 60 and 61: The model that we will use for the
- Page 62 and 63: Program 3dConvolve Command Line for
- Page 64 and 65: 2 Program plug deconvolve2.1 Purpos
- Page 66 and 67:
voxels indicated by the crosshairs,
- Page 68 and 69:
Deconvolution Plugin Screen Output
- Page 70 and 71:
The following Random Permutation an
- Page 72 and 73:
Shuffled array:0 5 6 1 3 5 0 2 1 1
- Page 74 and 75:
Program RSFgen Command Line for Exa
- Page 76 and 77:
Example 3.3.3.1 Initial Random Bloc
- Page 78 and 79:
Compare the output for this example
- Page 80 and 81:
Markov chain time series:0 0 1 5 0
- Page 82 and 83:
(X'X) inverse matrix::065 ;:000 ;:0
- Page 84 and 85:
Program 3dDeconvolve Screen Output
- Page 86 and 87:
then the standard deviation for the
- Page 88 and 89:
4 Program 3dConvolve4.1 PurposeAs t
- Page 90 and 91:
-input1DThe -input1D command specie
- Page 92 and 93:
-stim minlag k m-stim maxlag k n-st
- Page 94 and 95:
parameters b 0 and b 1 , and impuls
- Page 96 and 97:
Program 3dConvolve Command Line for
- Page 98 and 99:
Example 4.4.6 Single Stimulus No No
- Page 100:
5 References[1 ]F.Stremler, Fourier