In recent years, attitudes toward marijuana are changing and marijuana use is increasingly viewed as harmless. However, along the side of the increasing marijuana use there has been a rise in prevalence of DSM-5 cannabis use disorder (or CUD). This disorder is associated with adverse consequences such as cognitive decline, impaired educational or occupational attainment, impaired driving ability, emergency room visits, psychiatric symptoms and risk of addiction or substance use disorders. To learn more [1] , [2].
Here we examine possible mediators for developing CUD using (Wave 1) of the National Epidemiological Survey of Alcohol and Related Conditions (NESARC-III) dataset.
We take a partial correlation approach for the analysis of trios which may exhibit a causal relationship. These trios are modeled as partial graphs, in which information can flow in different paths: A->B->C, A->C, A->B, A->C->B. Where A is an "anchor node" - an event that clearly preceeds both B and C - and therefore cannot have entering edges.
.png)
In the given dataset, A can be an demographic feature (age or sex) or an early adverse event (EAE). Each time correlation is conditioned on a different node, blocking the respective information path, aming to revel the true stracture of the trio.
This code allows to examine mutiple trios and outputs a table which quatifies correlation reduction.
-
First you must obtain Wave1 NESARC-III dataset. In order to do so, you will have to request access as described Here.
-
Run Perprocess notebook for initial anylsis on the data and for the creation of the perprocessed set.
-
Run CUD notebook.
**logo created by IconMark from the Noun Project