In this work I have grouped stages into 'families' of related designs, and launch vehicles into families based on their first stage type. For instance, I consider the Thor Agena, Thor Able, and Delta II to be members of the same family, while Atlas Agena, Atlas Able, and Atlas Centaur are a second family. Ariane 1 to 4 and Ariane 5 are separate families - Ariane 1 to 4 share a common first stage design, but Ariane 5 is an entirely different rocket.
It's easier to maintain listings on a per-family basis, and the total number of first stage families is about 200 - much smaller than the 1000 or so different launch vehicle variants. There are some grey areas in assigning vehicles to families, but it's much less problematic than splitting hairs about slight differences between variants.
The rule I've adopted for defining a new family is a change in the tankage diameter. In some ways this is a superficial thing - to a rocket engineer, changes in the engine and the avionics will be much more significant than a change in the external dimensions - but in practise the diameter criterion works well and reveals important design commonalities between different launch vehicles. There are some marginal cases - the Trident II and Trident I missiles are closely related but have different diameters. The real complication comes in deciding which stage is the first stage. For orbital vehicles it's pretty clear, on a Delta II the Thor-derived first stage is a true first stage, it wouldn't make sense to count the GEM strapons as the core of the vehicle. But for many sounding rockets the vehicle consists of a booster and sustainer, with the booster falling off after only a couple of seconds. Is a Nike-Aerobee (Aerobee 170) a Nike variant or an Aerobee variant? After much agonizing I have decided to consider it a Nike variant. Similarly, Skylarks boosted by Cuckoo or Goldfinch motors are under the Cuckoo family, while single-stage unboosted Skylarks are in the Raven family.
The Family datafile contains: