I have to answer again, and again partially because of a misleading argument.
Let me state first that I think it is a matter of taste what kind of software one wants to use in research. If a researcher dislikes Excel, of course he can refuse to use it. The situation is slightly different for academic teachers. Not talking about spreadsheets and some of the statistics they can do when Excel is installed on practically every desktop in professional environments is dangerous because you are implicitly telling your students "we are so smart that we only are going to use software which is not accessible to dumber people". Not for your arguments (and I know Patrick's paper and I disagree with quite a few - but not all - of his points). -22 is indeed surprising, and people should be warned about it. But by itself it is not a reason to refuse to use it. If let Java compute 1234567788*2 you get -1825831720 No error message, no warning! I think this is much worse, and still we have rJava and many more R extensions based on Java. You also mention the possible loss of precision when you pass data through Excel. This is true if you use a character based transfer (e.g. csv-files). One of the reasons Thomas and I developed statconn and RExcel is that we wanted a transfer facility retaining precision. In fact, RExcel was written to overcome quite of the few shortcomings Patrick mentions in his paper. For academic teachers, I still think it is important to know the shortcomings of Excel and also know which ones can be alleviated by using RExcel. _______________________________________________ R-SIG-GUI mailing list [email protected] https://stat.ethz.ch/mailman/listinfo/r-sig-gui
