# 9.3. **How to get the degrees of freedom (DF) in an AFNI file?**¶

There are at least four ways to find out the degrees of freedom (DF) for a specific statistical sub-brick:

`3dinfo`

the AFNI GUI/viewer

`3dAttribute`

`p2dsetstat`

/`dsetstat2p`

The command line programs all apply equivalently to either volumetric or surface-format files. (Using the AFNI GUI just applies to the volumetric format.)

In the examples below, we use a volume called `stats.FT+tlrc`

. This
also happens to be the stats volume output by running the AFNI
Bootcamp demo example of `afni_proc.py`

in
`~/AFNI_data6/FT_analysis/s05.ap.uber`

. Assume that sub-brick #0 is
an F-statistic and sub-brick #8 is a t-statistic.

## 9.3.1. Using `3dinfo`

¶

The first method is to use `3dinfo`

, which outputs information about
the entire dataset. This can be piped through less, to be able to
scroll through it:

```
3dinfo -verb stats.FT+tlrc | less
```

On the screen you can find the header information about sub-brick #0
(the full F-statistic), just beneath the line starting ```
Number of
values stored at each pixel ...
```

:

```
-- At sub-brick #0 'Full_Fstat' datum type is float: 0 to 977.037
statcode = fift; statpar = 2 412
```

The two numbers under `statpar`

(=statistic parameters), 2 and 412,
on the second line are the degrees of freedom for this F-statistic.
(The statcode relates to internal code usage/labelling for identifying
this type of statistic; see the help of `cdf`

for a wee bit more
description, if intrigued.)

Further down you should be able to locate the other t-statistic sub-brick (#8):

```
-- At sub-brick #8 'V-A_GLT#0_Tstat' datum type is float: -11.3157 to 12.7985
statcode = fitt; statpar = 412
```

... and the DF count in this case is 412.

## 9.3.2. Using the AFNI GUI¶

Load the statistics volume `stats.FT+tlrc`

as an overlay in the AFNI
viewer (GUI).

Go through the following buttons:

`Define Datamode`

(just above the`Underlay`

and`Overlay`

buttons)`Misc`

(in the bottom row of the new panel that opens)`OLay Info`

.

You should be able to find the same information in the new window as
shown above with `3dinfo`

.

The same also works if you load the stats dset as an underlay, and
then go to `Define Datamode`

-> `Misc`

-> `Ulay Info`

.

## 9.3.3. Using `3dAttribute`

¶

For example, for the same sub-brick #0 above, run the following command:

```
3dAttribute BRICK_STATAUX stats.FT+tlrc"[0]"
```

... and you should get five numbers on the screen:

```
0 4 2 2 412
```

The last two are the DF for the F-statistic.

**NB:** For the sub-brick selection in the quoted squarebrackets, you
can use the label associated with the subbrick, which in this case is
“Full_Fstat”. So, the following would lead to the same output as
above:

```
3dAttribute BRICK_STATAUX stats.FT+tlrc"[Full_Fstat]"
```

Doing the same for the t-statistic sub-brick (with either the number- or string-label for subbrick selection):

```
3dAttribute BRICK_STATAUX stats.FT+tlrc'[8]'
```

... you will see:

```
0 3 1 412
```

... and the last number is the DF count for the t-statistic.

## 9.3.4. Using `p2dsetstat`

and `dsetstat2p`

¶

There are some complementary convenience programs for converting
p-values to statistics and vice versa, `p2dsetstat`

and
`dsetstat2p`

. These can be helpful for either scripting or for
seeing stats information (and they make use of the above
functionality, under the hood).

Similar to the `3dAttribute`

example, one looks at information for a
single subbrick. For example, querying information about subbrick #0
(or, again, one could use the string label, “Full_Fstat”):

```
p2dsetstat -inset stats.FT+tlrc'[0]' -pval 0.001 -1sided
```

... one obtains the output table:

```
++ Found input file : stats.FT+tlrc[0]
++ OK stat type : fift
++ BRICK_STATAUX : 0 4 2 2 412
++ params : 2 412
++ Final stat val : 6.3093
```

This might be more convenient than `3dAttribute`

for visualization,
because the “params” line highlights the one or more DF parameters,
rather than having to know *a priori* how many DFs to expect for a
given stat. Or, at least, this could be used to verify the
interpretation of the more opaque `BRICK_STATAUX`

info. **NB:** For
getting an appropriate statistic value, one has to use an appropriate
sidedness of testing (e.g., `-1sided`

for the F-stat here), but that
does not affect the DF info.

Running the similar program for the t-stat in #8, one might use:

```
p2dsetstat -inset stats.FT+tlrc'[8]' -pval 0.001 -bisided
```

... obtaining:

```
++ Found input file : stats.FT+tlrc[8]
++ OK stat type : fitt
++ BRICK_STATAUX : 0 3 1 412
++ params : 412
++ Final stat val : 3.3143
```

Running the complementary program `dsetstat2p`

will provide the same
DF info in either case. For example, going back to the F-stat in #0
(and using the classy label for sub-brick selection):

```
dsetstat2p -inset stats.FT+tlrc'[Full_Fstat]' -statval 3 -1sided
```

... yields:

```
++ Found input file : stats.FT+tlrc[Full_Fstat]
++ OK stat type : fift
++ BRICK_STATAUX : 0 4 2 2 412
++ params : 2 412
++ Final p-val : .025437950000000
```

On a final note for these programs, if you are really just wanting
either the stat or p-value output for scripting purposes, you can use
the `-quiet`

option, to just get that number and assign it to a
variable. In `tcsh`

syntax, this might look like:

```
set pval = `dsetstat2p -inset stats.FT+tlrc'[Full_Fstat]' -statval 3 -1sided -quiet`
```

... where subsequently using `echo ${pval}`

to display the value
yields:

```
.025437950000000
```