On 2013 Dec 20, Eric Johnson commented:

I get comments referring to this paper quite a bit, so I wanted to add a clarification that many casual readers seemed to have missed. The title uses "RADseq" generically, and then analyzes two different protocols: "RADseq" with the shearing step (Baird et al., 2008), and ddRAD which replaces the shearing step with a frequently-cutting enzyme (Peterson et al., 2012). The results are not the same for the two protocols. For example, Table 1 shows that "standard" RADseq deviates from the expected value of π by .995, while ddRAD deviates by .836.

All existing genotyping by sequencing methods that reduce the complexity of the genome will have biases and missing data when polymorphisms disrupt the sequence used to anchor the amplification and subsequent sequencing, and this paper shows how these biases can affect the calculation of population statistics. I just wanted to make clear the distinction between the two protocols, so when researchers use one or the other they can understand how to plan their projects and analyses accordingly.

[Conflict of interest note: I am an author on Baird et al.]

On 2013 Dec 20, Eric Johnson commented:

I get comments referring to this paper quite a bit, so I wanted to add a clarification that many casual readers seemed to have missed. The title uses "RADseq" generically, and then analyzes two different protocols: "RADseq" with the shearing step (Baird et al., 2008), and ddRAD which replaces the shearing step with a frequently-cutting enzyme (Peterson et al., 2012). The results are not the same for the two protocols. For example, Table 1 shows that "standard" RADseq deviates from the expected value of π by .995, while ddRAD deviates by .836.

All existing genotyping by sequencing methods that reduce the complexity of the genome will have biases and missing data when polymorphisms disrupt the sequence used to anchor the amplification and subsequent sequencing, and this paper shows how these biases can affect the calculation of population statistics. I just wanted to make clear the distinction between the two protocols, so when researchers use one or the other they can understand how to plan their projects and analyses accordingly.

[Conflict of interest note: I am an author on Baird et al.]