As mentioned in vignette 2], there are a number of options to explore the output of an analysis.
print(out) (or out): Returns an overview of all the settings of the model. This includes the number of participants and the formula used.summary: This will return a table that gives an overview of the statistically significant clusters per variable/term in the model. For each cluster, we report the number of vertices, the mean value of the vertex measure (e.g. thickness) in that cluster, the mean regression coefficient, and the mean standard error. Information on the location of the cluster can be obtained with summary(annot = TRUE), using the an annotation file from the target template (default: file = aparc.annot).Users may want to zoom in on specific variables or terms. To do this, we have to utilize “stacks”: stacks(out). This will return all the “stacks” in the model.
A stack is simply a term in the regression model. For example, a continuous variable like age will be stored as a stack called “age”. However, certain variables have multiple levels, and so each contrast needs to be stored separately. For example, a dichotomous variable, e.g. “disease” with levels “Yes” and “No”, will be stored in a stack called “diseaseNo”. A trichotomous version, e.g. “status” with levels “Control”, “MCI”, “AD”, will be stored in two stacks: “statusMCI” and “statusAD”. This has to do with the settings of how R deals with factors. By default, R transforms factors to become dummy-coded, which is reflected in the stacks.
By calling stacks(out), users can see which stacks were used in the model. Here are a number of functions that rely on stacks:
freeview(out, "statusAD"): Calls freeview (external window) for the statusAD stack. The freeview tool is a tool from the FreeSurfer suite that allows users to view maps of the brain.qdecr_snaps(out, "statusAD"): Calls freeview multiple times to generate multiple images, then compiles the images into one file. This function requires the magick package.Any analyses that are run with e.g. qdecr_fastlm are automatically stored as files on disk.
To load the project back into R, you can use the qdecr_load function:
vw <- qdecr_load("path/to/project")
It is also possible to reload the vertex-wise data as vw$mgh:
vw <- reload(vw)
Do note that this will generate huge temporary files in your dir_tmp, so be mindful to remove those afterwards!
Users can also load (and save) individual mgh files. Three functions are currently in use:
load.mgh("path/to/mgh/file"): Loads an .mgh file into R as a list, where the x element contains the vertex-wise values.as_mgh(x): Converts a vector x to an object that resembles an .mgh file.save.mgh(x, "path/to/mgh/file"): Given that x resembles an .mgh file, it will save x to the specified path.