An R implementation of the Necessary and Probably Sufficient (NPS) test for finding valid instrumental variables, as suggested by Amit Sharma (2016, Working Paper) <http://amitsharma.in/pubs/necessary_probably_sufficient_iv_test.pdf>. The NPS test, compares the likelihood that a given set of observational data of the three variables Z, X and Y is generated by a valid instrumental variable model (Z -> X -> Y) to the likelihood that the data is generated by an invalid IV model.