Hey Gang,
I just tried a new PATH analysis with more ROIs, and I have noticed that when using the covariance matrix for model validation and tree growth if i specify a reciprocal connection using my theta matrix then I get the extremely large values. However, if the path I have identified is non-reciprocal i get a value that is interpretable. Is there maybe something to do with reciprocal connections and rapid-event design that is maybe influencing the large connection coefficients? Below is the my example for model validation with a covariance matrix:
1dSEM -theta thetas.1D -C SEM_cov.1D -DF 747.412 -psi SEM_resvar.1D -limits -10 10
++ 1dSEM: AFNI version=AFNI_2009_12_31_1431 (Mar 17 2010) [64-bit]
++ Authored by: Daniel Glen, Gang Chen
Initial Theta Setup Matrix matrix: 5 x 5
# amy dlpf pgac temp temp
amy 0.0 0.0 1.0000 1.0000 1.0000
dlpfc 0.0 0.0 0.0 1.0000 1.0000
pgacc 0.0 0.0 0.0 0.0 0.0
tempA 1.0000 1.0000 0.0 0.0 1.0000
tempP 1.0000 1.0000 0.0 1.0000 0.0
++
++ Connection directionality is from column to row
++ Finding optimal theta values
++ Total number of iterations 142615
++ Cost is 52.0284
++ Chi Square = 38834.6
Connection coefficients matrix: 5 x 5
# amy dlpf pgac temp temp
amy 0.0 0.0 -0.5286 -9.9998 9.9998
dlpfc 0.0 0.0 0.0 9.9997 9.9997
pgacc 0.0 0.0 0.0 0.0 0.0
tempA -9.9998 -10.0000 0.0 0.0 -9.9648
tempP -9.9995 9.9997 0.0 9.6834 0.0