Wed Oct 29, 2003

Number Crunching

Kevin Drum is keeping score in an argument about data on global warming. “M&M” (don’t ask me, I’m only reporting this) re-analyzed data for a famous graph and claimed to find serious errors. Now, Kevin says

Somebody it’s not entirely clear who exported the original raw data to Excel but somehow exported 159 columns of data into a 112-column spreadsheet. M&M failed to compare the spreadsheet to the original data and thus produced a “correction” that was riddled with errors.

Here’s something you can try at home: Walk up to a statistician, shake their hand and say “When I reanalyzed your data using Microsoft Excel, I found numerous errors.” Stand well back. Wait.

It’s all the worse, really, given that one of the best pieces of software for statistical computing is available for free. In fairness to these guys, though (and in response to a comment below from Kevin), I should say that data management is an often error-prone business that I’ve been bitten by myself. It’d be tough on them if an otherwise well-conducted reanalysis got tripped up because they used an incorrect version of the dataset.