Is it possible to set a random seed - each run of chicago yields a different number of interactions

Create issue
Issue #23 new
Former user created an issue


Is it possible to set a random seed for CHiCAGO? - each CHiCAGO run yields a different number of significant interactions (score >5) due to low scoring interactions sometimes slip over the threshold of 5, I guess this is because of the random sampling used to build the model?. Currently i've been dealing with this by running CHiCAGO several times and then selecting the shortest list of interactions to take onto further analysis as that list should have the least false positives? Is this the most appropriate way to deal with this or is there a better solution?

Many thanks Charlie

Comments (1)

  1. Mikhail Spivakov

    Thanks for this suggestion. We will think of making this option user-friendly, but so far it's already available if you execute chicagoPipeline() in a custom script (as opposed to as part of runChicago.R). In this case you can simply set the seed immediately before invoking this function. (See the Chicago vignette on the sequence of functions to run for full analysis - it's very straightforward).

  2. Log in to comment