Basic question about value in olay panel

AFNI Message Board
1

AFNI version info (afni -ver): 23.1.10

Hi.
I'm a begineer, and also my english skill's quite low so please forgive me.

I making a overlay .nii.gz file which contain BOLD signal and temporal data through 3dttest++ command only subburick option.However, I didn't get the meaning of value 'The number in the top and bottom right' in this page(5.1.5. AFNI GUI: Overlay panel — AFNI, SUMA and FATCAT: v23.1.10).

The bootcamp's video([AFNI Academy] AFNI GUI: View model results (olay=coef, thr=stats) - YouTube)is right, the meaning of Coef sub-bricks 'The number in the top and bottom right' is BOLD signal change rate, right? But, I think my data's value is quite low, and I can't get enough understanding about it, so I wirte the question.

I'm so sorry about basic question.
Thanks.

code text  # or delete if not needed
2

Hi-

When you run a ttest, you will have multiple pieces of information stored in the resulting dataset (which seems to be a NIFTI volume here). For example, you get the mean value or difference, and you will get an associated statistic, either a t-stat or Z-score depending on what you wanted. If you run:

3dinfo -verb DATASET_STAT

you will see the labelled list of what each subvolume (or "sub-brick") contains. For example, in a Bootcamp example of a paired t-test output, this is what part of hte 3dinfo output looks like:

Number of values stored at each pixel = 6
  -- At sub-brick #0 'Vrel-Arel_mean' datum type is float:     -7.38705 to       8.88611
  -- At sub-brick #1 'Vrel-Arel_Tstat' datum type is float:     -9.56758 to        11.147
     statcode = fitt;  statpar = 9
  -- At sub-brick #2 'Vrel_mean' datum type is float:     -10.7478 to       10.2102
  -- At sub-brick #3 'Vrel_Tstat' datum type is float:     -13.4342 to       13.2141
     statcode = fitt;  statpar = 9
  -- At sub-brick #4 'Arel_mean' datum type is float:     -9.00484 to       10.2745
  -- At sub-brick #5 'Arel_Tstat' datum type is float:      -16.952 to       12.9209
     statcode = fitt;  statpar = 9

There were 3 contrasts examined, so there are 3 pairs of "mean value" (like in sub-brick 0) and "statistic value" (like in sub-brick 1, which is the t-stat associated with the mean in sub-brick 0---both have the same prefix, "Vrel-Arel", at the start of their name, as well as being stored in sequence).

So, if I wanted to display one of these effect-and-statistic pairs in the AFNI GUI, I would select, say, sub-brick 0 to be the Olay (because this has the units of whatever I was investigating, and is the effect from my comparison) and sub-brick 1 to be the Thr (because the statistic information describes the uncertainty, and so I can use it to higlight the more "certain"/significant regions). I could also put subbrick 2 as Olay and then I would use subbrick 3 as Thr, etc.

In general, showing the effect estimate (the coefficient, beta weight, or in this case 'mean') as Olay makes sense--you can set hte colorbar range appropriately based on the units of whatever that has. In FMRI, this might have units of "BOLD percent signal change" in many cases, such as if you processed the data with afni_proc.py and used the 'scale' block; if you have additional covariates or further model complexity, units can change, too, in some cases. Then, you can include the statistic information about relative uncertainty by putting it as the Thr (the threshold), and highlight where the effect seems quite high relative to its uncertainty, essentially.

As a further note, the case for using transparent thresholding with the A and B buttons is made here:

... and here:

--pt

1 Like