2 In practice it is rather difficult to identify details you to definitely predict migration but not companion particular. We analyzed the newest appropriateness in our model specifications pursuing the statistical take to recommended by the Leung and you will Yu (1996) and you can Puhani (2000). It believe the amount of collinearity between the explanatory regressors forecasting companion sorts of and you can taken from quoting the likelihood of Durham quarters (equation (2)) is the definitive mathematical standard to gauge the fresh new appropriateness of your own model requirements since it grabs the brand new extent to which the latest analytical requirements separates the 2 behaviors. Particularly, if status number of the brand new matrix of your regressors (this new inverse Mills proportion regarding formula (2) forecasting partner type) is higher than 20 then the design is not sturdy therefore the crucial details are not splitting up the probability of migration out of brand of companion.