On Thu, May 5, 2016 at 10:33 AM, Matthew Brett <[email protected]> wrote: > On Thu, May 5, 2016 at 9:45 AM, Waldman, Simon <[email protected]> wrote: >>> -----Original Message----- >>> From: C. Titus Brown [mailto:[email protected]] >>> Sent: 05 May 2016 14:41 >>> To: Waldman, Simon <[email protected]> >>> Cc: [email protected] >>> Subject: Re: [Discuss] Word and PowerPoint "all wrong"? >>> > > The fact of the matter is, Excel has been demonstrated time and >>> > > again to be not just inefficient for scientific analysis but usually >>> > > out-and- >>> out wrong. >>> > >>> > [citation needed] >>> >>> Here's one that has had a reasonably significant impact on biology: >>> >>> http://bmcbioinformatics.biomedcentral.com/articles/10.1186/1471-2105-5- >>> 80 >> >> Interesting, thank you. I'm not sure that's so much "Excel is wrong" as >> "Excel was allowed to make a guess at a data type, and guessed wrong" - but >> it certainly has the same effect for the unwary. Worth being aware of. >> > > This is a list of errors in statistical calculations in Excel, with citations: > > http://statisticalengineering.com/Weibull/excel.html > > But, more important than that, is that Excel makes it easier to make > errors and harder to detect errors. There's the famous Reinhart and > Rogoff case, as just one example: > > http://www.peri.umass.edu/236/hash/31e2ff374b6377b2ddec04deaa6388b1/publication/566 > > One error was a simple drag-select mistake. > > Having watched people use Excel for many years, it's surprising to me > that these errors don't happen more often than they do. I remember > being terrified at my colleague's explanation of the work she was > doing by column dragging and and spreadsheet tab manipulation, and I > knew her to be a very careful scientist. > > I wouldn't worry too much about that, if I did not have a very clear > idea of my own ability to make silly mistakes, I lesson I largely > learned from testing my own code [1]. > > So, if I have data or analysis where the potential for error is low, > or the cost of error is low, then I would consider using Excel. If I > want to be able to track the entire calculation, read it, debug it and > test it, I need a tool that is designed to do that.
On a meta-level - that is the one thing that I wish I could teach my students - that they make mistakes all the time. All my current practice is motivated by realizing that is true. It took me a long time to realize that was true. I wouldn't care nearly as much about which tools I was using, if that wasn't true. That's a question for SWC as an organization. Is it teaching to make science more efficient, or to make science more correct? Of course the answer will be 'both' but the emphasis will be different, depending on which we care about more. More efficient, then you might emphasize tools that can get a lot of analysis done in a short time. More correct, you might emphasize tools that push you to think harder about the problem, at the expense of a steeper learning curve and less analysis per unit time. Best2, Matthew _______________________________________________ Discuss mailing list [email protected] http://lists.software-carpentry.org/mailman/listinfo/discuss_lists.software-carpentry.org
