What does the "Intensity" computed by FIM mean?
Document Actions
Q35. What does the "Intensity" computed by FIM mean?
The "FIM intensity" is defined on page 22 of the manual afni200.ps. It is the parameter alpha in the least squares fit of each voxel time series x(t) to the model
x(t) = alpha * r(t) + a + b*t + noisewhere r(t) is the reference waveform you supply. Thus, if you double r(t), alpha will be halved; alpha is the amount of r(t) present in x(t), as computed by linear regression.
The "Compute FIM+" menu item in an AFNI graph window has the ability to compute percent change in addition to the fit coefficient alpha. In this module, "percent change" is defined from the baseline in the absence of function, not from the average time series value including the function. r(t) is linearly scaled and shifted to be between 0 and 1 (no matter what the input range actually is), and then the fit above is computed. The baseline is a + b*tmid, where tmid is the middle time point, and the percent change is 100*alpha/baseline. (If other orts are present, baseline is defined as the average value of the fit of all the orts, including the implicit orts 1 and t, whose coefficients are a and b in the equation above.)
Note that this will often produce some huge percent changes of low significance outside the brain (where the baseline is small), so that the range of the "% Change" sub-brick will be large (cf. Q49). For proper color overlay display of 5% (say) signals, you will have to turn off the autoRange toggle on the "Define Function" control panel, and enter the desired range in the parameter box directly below.
Finally, note that the -percent option of 3dfim does something slightly different: it computes the absolute value of the percent change from the time series mean (which will include the mean value of the function). Why is this different from the "FIM+" module within AFNI itself? There is no good reason. (See also the answer to Q41.)
This FAQ applies to: Any version.
