The statistical uncertainties in a 1000 year northern hemisphere mean temperature reconstruction obtained from 17 proxy chronologies are examined in detail by analysing the range of estimates obtained from all possible subsets of the proxy collection with up to 6 proxies omitted. The study is motivated in part by the large range of recently published reconstructions in the 15th and 16th centuries. The analysis is also applied to a previously published version of the reconstruction, based on 13 proxies, to investigate the increase in estimated skill generated by the 4 new proxy series. The reconstructions are also evaluated against estimates of 1750-1850 temperatures based on entirely independent data. |