Woodward and Bernstein were pikers compared to Steve. What persistence.

Steve may just be listed in a future edition of “Famous Auditors in History” as the originator of a concept known as McIntyrian Long Term Persistence (MLTP).

Again, please don’t assume that this has the effect of ignoring “valid” proxies. For all we know, the entire proxy network is just a garbage bin.

Specifically with regards to detecting the Little Ice Age via proxies?

I don’t think that the main issues really pertain to multivariate mechanics, but to the inconsistency of “proxies”. If every time series was temperature plus some sort of low order red noise, then almost any method would extract the underlying signal. The problem arises when you don’t know that your “proxies” really are temperature plus low-order red noise.

The sort of evidence that would most impress me at this point is the determination that some class of proxy really is a “good” proxy. And then not using any of the data used to decide that it was a “good” proxy, but getting fresh samples yielding the same result. Right now, what does any of this mean if Graybill and Ababneh get radically different bristlecones; and Briffa throws out the Polar Urals extension.

The actual temperature history is calculated using all available global temperature records, including the regions from which no proxies were used in the study. I presume that you are suggesting that if a set of weights {wk} (for example, by gridbox area) is used in determining the global average, then the unrepresented regions are removed and the remaining ones re-weighted proportionally to their original weights. However, some regions may warm faster or slower than others due to other climate factors. Could this not possibly introduce biases into the situation based solely on the fact that proxies were or were not selected from those regions?

Several alternatives come to mind:

Expand the regions represented by given proxies based on distances between locations of the proxies before calculating area weights.

Calculate a “global record” to compare to based solely on the regions which have proxies.

Calculate a relationship between measured temperatures of the proxy-represented regions and the average global temperature (of all the regions) to determine the weights. This was what I was suggesting in my earlier post.

I am not sure that I completely agree with you on this point:

By the way, if you calibrate a local series to temperature and then do your gridding and averaging, you should be done. You should not rescale it again against the hemispheric average

It is quite possible that particular geographic regions could be better or worse at reflecting global temperature conditions. Thus, giving higher emphasis to proxies from the “better” regions might give improved results. I agree strongly with you that what shouldn’t be done is to reuse the proxy data to determine the weighting scheme. Instead, the weights need to be determined solely from the measured temperature records from each of the regions. For example, simple (or even inverse) regression could be a means for “calibrating” the local temperatures to global scale. This allows for what I see as a more accurate representation of the physical situation: the proxies provide information on the local conditions and the actual local conditions give further information on the global situation. Re-using the proxies in the final stage bypasses the second step by linking the proxies directly to the global record through what can only be justified by relying on some unknown, scientifically unexplained, teleconnection mechanism.

By the way, if you calibrate a local series to temperature and then do your gridding and averaging, you should be done. You should not rescale it again against the hemispheric average

Yes. I whined about this earlier,
http://www.climateaudit.org/?p=4038#comment-303777