Is the earth getting warmer, or cooler?

random versus systematic error.

Quoteth:From a statistical viewpoint, data recalculation should cause each year to have a 50/50 probability of going either up or down - thus the odds of all 70 adjusted years working in concert to increase the slope of the graph

This statement is only valid concerning random errors. Such errors are quantified by statistics and are shown by error bars. All data has random error, and most properly reported data includes an indication of random error, most often in the form of error bars. Certainly one does not flip a coin to choose which side of the random error one will choose for each point. It is random, it is a confidence level, so all that can be done is to state the theoretical error.

There is a second type of error. This is systematic error. This error occurs not by random chance, but by some material error in the measurement process. Systematic error can be caused by a bad reading, an improper zero offset, or a bad calibration curve. Systematic errors can sometimes be detected through statistical analysis, for instance one might discard all 'outliers', at one's own peril, but most often these errors either remain undetected, or, upon detection, require exactly the kind of recalculation condemned in the article.

In the present case, there is no credible reason presented in the article to indicate that some older data might be preferentially treated with respect to some new data. Certainly, one can imagine that older instruments might have systematic errors that are different from those on newer instruments. We can even suggest that an analysis of the random error on older instruments might be such that the states skewing of the graph might occur.

I am not saying that the NASA analysis is correct. However, the authors argument seems sufficiently invalid to be ignored.


Back to the forum