In talk.politics.drugs Erkki Komulainen <[EMAIL PROTECTED]> wrote: > :> 'rank-transformation' to the data; and whether that is > :> desirable or useful or damaging.
> : By that you are turning scores into ranks? You *have* to do that when the > : data is non-linear is my guess. Then you put it into a formula designed to > : work with ranks. Otherwise what you get out has no meaning. > Hi! Ranks or raw - the problem of non-linearity remains the same. If you have have linear data it would be silly to use ranks. You owuld lose power. I guess the idea of product moment is a bit like balancing a see saw. You are multiplying a weight times the distance from the balance fulcrum and need equivalent calculations. So if mum is 3 times as heavy as toddler she will have to be 1/3 distance from the fulcrum. That works in that case because of linearity. But take the example of the shouting competition. A voice with twice as much power will not be heard twice as far away. The relation is not linear. But you can still say that the louder voice will be heard further away - things are still ranked. If the see saw balance were like that you could not balance by knowing weight and calculating linearly. But you could still get somewhere by putting mum a bit closer to the fulcrum than toddler. So if you have linear data, keep it as scores, don't waste the power. So your statement really is null. Examine > the biplot scatter visually and put a lowess the what is that? to help to see how the > relationship deviates from a linear one. It often helps. I should hope so, indeed. Where is the training in this? BTW ranks and raw > correlate in my experience often above .9 - even close to unity when > there are lots of ties. With what sample size? It is also possible to get a reasonable approximation when adding up money if amounts of cents 50 or over are rounded up to the next dollar, amounts of cents below 50c are rounded down to the lower exact dollar. By the tiem you are up to $30 you won't be a great percentage off. But you could be quite a percentage off at $10. Also you might get rather out in a shop which sells everything at some dollar and 49 cents. -----= Posted via Newsfeeds.Com, Uncensored Usenet News =----- http://www.newsfeeds.com - The #1 Newsgroup Service in the World! -----== Over 80,000 Newsgroups - 16 Different Servers! =----- . . ================================================================= Instructions for joining and leaving this list, remarks about the problem of INAPPROPRIATE MESSAGES, and archives are available at: . http://jse.stat.ncsu.edu/ . =================================================================
