An acquaintance of mine who’s very statistically savvy (and quite conservative) posted the following link on Facebook today.
I replied as follows (I’ve replaced a link here with a clickable image):
As a statistics guy, you know way better than most how important sample size is. There was a 30-year plateau in the HADCRUT data, mid-40s to mid-70s. But the longer-term trend is obvious and apparent:
(This is one just data set, but you can view and compare many others at this link. They’re remarkably similar. Try the rolling-average smoothing to get a less noisy picture.) For me the really big sample relates to arctic ice. We don’t really know when the summer arctic ice cap was last as small as it is now, but we do know that the last time it melted completely was approximately three million years ago. To me, if that’s just sort of happening due to random climate variation, over mere decades, it’s looking like a quite spectacular statistical anomaly.
I added: if he’s been saying this for years, he’s been basing those statements on even smaller sample sizes.
Cross-posted at Asymptosis.