scatter charts and bar charts after group level analysis

AFNI Message Board
1

Hi，AFNI Group

For analyses of effects of illness duration, medication, episode, and symptom severity, we calculated a linear mixed-effects models (LMM) by using 3dLMMr in AFNI: y ∼1 + HAMD + Episode + Medication + Duration + Education + Age + Sex + Motion + (1| Site). This LMM describes the relationship between a response variable (e.g., VMHC) and independent variables (here, HAMD, Episode, Medication, Duration, and covariates of age, sex, education, and head motion), with coefficients that can vary with respect to the site.

By setting a threshold of p<0.001, we found significant HAMD effect and Episode effects(i.e., 2 clusters: Clust_HAMDEffect_p0.001_10_mask+tlrc and Clust_EpisodeEffect_p0.001_10_mask+tlrc). The current reviewer wants us to show HAMD effect with scatter charts and Episode effect with bar charts.

How to deal with this problem effectively? Can the Cluster Explorer feature in AFNI handle this? Maybe there’s a better way? Thanks!

PS: I have: (more specifics below):

  1. An output dataset from 3dLMEr.
    （i.e., Clust_HAMDEffect_p0.001_10_mask+tlrc
    and Clust_EpisodeEffect_p0.001_10_mask+tlrc).

  2. All of my individual subject datasets.
    i.e., 2000 VMHC maps that went into the group level analysis (.nii).

  3. A table(.txt) with my subject ID, HAMD, Episode, Medication, Duration, Education, Age, Sex, Motion, Site, InputFile(xxx.nii).

  4. The master template to which the input datasets were aligned (i.e., ch2bet.nii.gz). This will serve as an underlay for viewing results.

2

The current reviewer wants us to show HAMD effect with scatter charts and Episode effect with bar charts.

Unlike in a simple model, scatter plots are not really revealing in this case within an LME framework for two reasons:

  1. The involvement of a few covariates make the associations quite complicated.
  2. Under a multilevel (or mixed-effects) model, the difficulty of visually presenting the hierarchical data structure in a practically revealing manner becomes overwhelming in illustrating the data stratification across subjects. See an example below:

http://doingbayesiandataanalysis.blogspot.com/2019/07/shrinkage-in-hierarchical-models-random.html