From the above quoted wikipedia page
http://en.wikipedia.org/wiki/Redshift_quantization
Quote:
|
"The above an intrinsic redshift hypothesis, if true, will have far-reaching consequences for cosmology and the nature of QSOs. Most of those previous studies on the Karlsson formula used rather small samples (except for Arp et al. 2005), and have been suspected that the claimed peaks were due to artifacts associated with selection effects (Basu 2005). To avoid such a heterogeneous selection manner as well as personal prejudice, Hawkins et al. (2002) tested the periodicity in log(1 + zqso) with 2dF redshift survey data with 67291 nearby galaxies and 10410 QSOs; it was found that there is no periodicity in log(1 + zqso). However, Napier & Burbidge (2003) argued that in order to use the 2dF sample to properly test the original hypothesis, it is necessary to establish for each pair that the galaxy is at least a late-type active spiral system. Arp et al. (2005) also re-examined the 2dF sample and claimed that they found that the redshifts of brighter QSOs in the QSO density contours fit very exactly the long standing Karlsson formula and confirm the existence of preferred values in the distribution of quasar redshifts."
|
I want to explain why, if Arp is right, the Hawkins result is expected to be null and the Arp result is successful. According to Arp, QSOs are not at cosmological distances implied by redshift but have two componets to the redshift that he calls "internal" and I guess "external" or cosmological. The periodicity in QSO redshifts is in the internal component so that it shows up in samples that are nearby (all at small z compared to the periodicity in log(1+z)). The Hawkins result includes QSOs which are distant and therefore have large cosmological redshifts. This will totally blur out all the peaks in the internal redshift. When Arp uses only bright QSO then he effectively selects the nearby ones that all have small external z component so no blurring results.
The problem is that when analysis is done to compare two models, you cannot make the assumptions of one model when testing the other. Unfortunately these procedures are so ingrained that it happens without people realizing.