Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.
From | jpitblado@stata.com (Jeff Pitblado, StataCorp) |
To | statalist@hsphsun2.harvard.edu |
Subject | Re: st: -anova- command fails with mildly complexx model |
Date | Fri, 16 Apr 2010 17:53:36 -0500 |
We promised more information about the bug in -anova- reported by Al Feiveson <alan.h.feiveson@nasa.gov> that caused -anova- to misinterpret some specifications and sometimes even to crash Stata. Quick update ------------- We have fixed the problem and that fix is now in testing and certification. We expect to have it out as an update in the middle of next week. Longer update ------------- One of Al's problem commands was, . anova y treat group treat#group / isub|treat#group time time#(group##treat) The crash is triggered only when parentheses are used to bind terms in a factor varlist and when interactions or subsets of interactions among the bound terms also appear elsewhere in the specification, such as Al's -treat#group ... time#(group##treat)-. The bug is not limited to -anova-, it can occur in any command that accepts factor variables. While waiting for the update next week, Al can run his model by explicitly specifying his terms, rather than relying on parentheses expansion. If instead of typing, . anova y treat group treat#group / isub|treat#group time time#(group##treat) Al types, . anova y treat group treat#group / isub|treat#group time time#group time#treat time#group#treat his model will estimate correctly. -- Jeff Pitblado -- Ken Higbee -- Vince Wiggins jpitblado@stata.com khigbee@stata.com vwiggins@stata.com <end> * * For searches and help try: * http://www.stata.com/help.cgi?search * http://www.stata.com/support/statalist/faq * http://www.ats.ucla.edu/stat/stata/