Today was get down to business day. I began messing around a bit with the Because package to generate my own synth() files, or artificial causal models. The process is a little complex at first glance, but once you get used to it – super cool. I made up some brain biomarker data (I used to be a neuroscientist, sue me) based on linear relationships, along with a few confounders thrown in for fun. I was able to use Because to establish the causal probabilities between each variable, as well as the Bayesian likelihoods/probabilities of each variable coming out to be a particular value or set of values. Using cdt I was able to run a PC algorithm on the model, and it in turn was able to give me its predicted relationships that I could compare with the original ground truth. Unfortunately I couldn’t get the visualizations to work to show the DAGS that I wanted, but I’m hoping to get that solved soon! I’ve also been trying to get LUCAS thrown into the mix since it’s a pre-generated, complex and binary model (looking at you, UCM algorithm…), but that’s a work in progress as well.
Searching For a Cause 