Arp has provided a number of high redshift objects and low redshift objects which are clearly joined together and are part and parcel of the same thing. Arp is not the one in denial here.
I can show you more than a couple of papers that show that statistically, given the number of quasars in the universe, and given the number of normal galaxies, you can show that the number of quasars and normal galaxies that are seen to "attached" are not statistically any more than those predicted through the number densities of quasars and normal galaxies.
Realistically, for quasars and galaxies to be dynamically related, especially in Arp's theory, quasars should be preferentially clumped around normal (or even active) galaxies. They are not. Not even a little. Both distributions, on all angular scales, are isotropically and homogeneously distributed throughout the universe. No clumping at all. This convincingly shoots Arp's theory down.
I think the Sloan Digital Sky Survey will be redoing this statistical analysis, just for grins, since it will be the best quasar survey ever done to great depths in redshift.