GraphVar Turbo GLM Tutorial
GraphVar Turbo GLM Tutorial
Outline:
m: Mean
significant
not-significant
Hypothesis:
• Research_site and sex are initially not selected as these variables are encoded as
strings in the variable spread sheet … you may select them if you want
• Simply select all the thresholds in the box with „ctrl+A“ (see manual for how to add more thresholds)
• For this investigation let´s construct subject specific null-model networks to calculate „small-worldness“
• Here, we ONLY generate 10 binary random networks per subject per threshold using the
„randomizer_bin_und“ BCT function
(for small-worldness, normalization purposes, or non-parametric testing you would normally use 100-1000 or even more …
but this will take a lot of time)
• As we have the hypothesis, that chilli eating champs probably have more efficient brains and that
probably insula and orbito frontal cortex may contribute here, we select:
• FYI: you can also add custom functions (see appendix in the manual); also note that for some of the
functions it would not make sense to do statistics on (e.g., modularity affiliation vector; get components)
• In the GLM panel add „eating_contest_chilli“ and „beer_pong_score“ as predictors in the
between covariates field
• Also select the option to perform permuation testing with 1000 permutations per threshold
• If you have the parallel computing toolbox installed, you may want to use more workers (cores) to
speed up the generation of null-model networks!
• Also use the colour map editor to set the range of correlations in the clolour bar (e.g., -1 to 1)
• We think that these results are meaningful and decide to save and to export
these to a csv file (which we open with excel later on)
• Only things that are visible in the results window will be exported
Everything we have computed (global efficiency and local efficiency across thresholds) may be saved
Interpretation
Hypothesis – confirmed!
Within Variables
Within Variables