Post Processing
Analyses performed on CUBIC:
- ModelArray (version 0.0.0.9000); ran linear model on ASL data (n = 209; 5 participants from the full 214 with processed ASL were removed after QA checks - basically those with QEI below 0.6 and rmsd over 0.3 were manually inspected for quality, those who did not pass manual inspection were excluded. Analyses and related files at
cbica/projects/GRMPY/project/curation/testing/stats_ModelArray
).- Followed documentation from Model Array demo and quick start guide
- Used Confixel for conversion from .nii to .h5; final convoxel .csv available at
cbica/projects/GRMPY/project/curation/testing/stats_ModelArray/GRMPY_convoxel_meanCBF_n209_finalversion.csv
- Ran
qsub_call.sh
, which calledrunRscript.sh
which called the actualRScript.R
. This used the singularity imagecbica/projects/GRMPY/project/curation/testing/stats_ModelArray/modelarray.sif
- R script used the formula
CBF ~ age (negligible non-linear effects) + sex + basilQEI + rmsd + als_score/bdi_score/scared_score_total/ari_score
and function as below:
mylm <- ModelArray.lm(formula, data = modelarray, phenotypes = phenotypes, scalar = "CBF", element.subset = element.subset, full.outputs = full.outputs, correct.p.value.terms = c("fdr", "bonferroni"), correct.p.value.model = c("fdr", "bonferroni"), pbar = TRUE, n_cores = num.cores) #n_cores = 4, element.subset = all voxels, full.outputs = TRUE
- Converted from .h5 to results folder named after date, containing all niftis, again using Confixel.
- No significance found for clinical scales: including fdr_correction images, cluster correction via
easythresh
and viewing viamricron
.
Results directories are pointed to in the file
ReadMeResultFolderDescriptions.md
in thestats_ModelArray
directory.Results further confirmed by vector testing in R:
GM (GMmeanCBF from ASLPREP_QC.csv) ~ age + sex + basilQEI + rmsd + als_score/bdi_score/scared_score_total/ari_score
- Did you use pennlinckit? Not yet.