With that being said, I doubt that survivorship bias is strong enough to explain their combined result of 10/17 models being significant.
> “We conducted a literature search to identify papers published prior to the early-1990s that include climate model outputs containing both a time-series of projected future GMST (with a minimum of two points in time) and future forcings (including both a publication date and future projected atmospheric CO2 concentrations, at a minimum). Eleven papers with fourteen distinct projections were identified that fit these criteria.
This brings to mind the old joke: "I used to smoke weed. I still do, but I used to, too."
Specifically, the phrase "were identified that fit these criteria" sets off my skepticism spider senses. Kind of like the popular "97% of scientists..." meme, that doesn't quite hold water when you dig under the covers.
To be clear, I'm not saying deceit is taking place, I am simply saying that I am skeptical.
I wasn't able to find any details on the specifics of the literature search, such that the search could be reproduced. Any idea if that was included in the study and I somehow overlooked it?