Dennis Roberts wrote: > > well, i am not convinced that stevens and others were all that misguided > ... have a look at this "heavy reading" paper by warren sarle of SAS > > it all has to do with what kinds of statements you want to be able to make > about your analyses > > ftp://ftp.sas.com/pub/neural/measurement.html
An interesting article, though I don't necessary agree with all of it (e.g., he finds Michelle's arguments unconvincing, but I'm not convinced they can be discounted so easily). Also, I think there are perfectly reasonable and common situations where 'ordinal scales' can be treated as interval plus error for practical applications (and in fact there may be some statistical basis for this if certain conditions are met). He also squashes some standard misconceptions about binary measures and discrete measures being problematic for standard parametric stats. Thom . . ================================================================= Instructions for joining and leaving this list, remarks about the problem of INAPPROPRIATE MESSAGES, and archives are available at: . http://jse.stat.ncsu.edu/ . =================================================================
