Usually, research make use of poisson regression models as the

number of patents is characterised as a count variable which in turn is considered to have a non-negative discrete nature and an excessive number of zeros.

As triadic patent family data contains fractional count data, however, a poisson regression is not feasible in this case.