Cook 2016 includes 14 previous studies, and omits 2. For 10 of the previous studies, Cook 2016 shows the consensus rate including don't know / no position. For 3 (Cook 2013, Verheggen, Rosenberg), Cook 2016 excludes don't know / no position. For 1 (Oreskes) Cook 2016 shows the consensus rate excluding no position, and the sample size including no position. Following Cook's majority position, I changed all results and sample sizes to include don't know / no position.

For the Carlton study, Cook 2016 copies a small error in that paper, and inflates the sample size from 38 to 306.

For the Stenhouse study, Cook 2016 changes definition. In the other studies, agreement is with the hypothesis that humans are responsible for more than half of the observed warming. Although Stenhouse reports the rate of agreement with this hypothesis, Cook 2016 replaces it with the weaker hypothesis that humans contributed to warming.

The graph below shows the impact of this lack of consistency. In black, it shows the rate of consensus as estimated in the literature and as reproduced by Cook. In gray, it shows estimates omitted by Cook. In red, it shows estimates that were replaced by Cook. In green, it shows the replacements.