🚀 go-pugleaf

In article <1992Oct29.163032.11967@egh-qc.uucp> pete@egh-qc.uucp (Pete Phillips) writes:

>I have recently read a paper where the authors carried out a
>"Multidimensional Space Analyses".

>Essentially the authors were comparing 6 different aspects of a
>product, and used a formula where they took the square root of the sum
>of various parameters squared.  A low index indicated a good product
>and a high index a bad product.

>Are there any good references on this, and is it a well recognised
>method of producing a single index figure from a number of different
>parameters ?

That is, the authors derived the euclidean distance of the product from an "
origin" in the space of the 6 product attributes.  It's just Pythagorus' 
theorem extended to 6 dimensions. (An obvious assumption in such an analysis 
is that the product attributes are orthogonal - otherwise they should have 
calculated Mahalanobis distances which take into account the correlations 
amongst attribute vectors.  Of course the attributes could be Principle 
Components derived from a larger number of attributes.)

Intuitively, a "small" index, or Euclidean distance, would require that the 
parameters were relatively small and/or of equal length.  Whether that 
constitutes a "good" product I've no idea.  But I'd be very interested to 
know of other opinions.

  - - - - - - - - -
 |      _--_|\     |   Hume Winzar
 |     /      \    |   Commerce School,
 |     *_.--._/    |   Murdoch University,
 |           v     |   Perth, Western Australia
  - - - - - - - - -  
                       E_Mail winzar@csuvax1.csu.murdoch.edu.au
                       Phone: (09) 310 7389
                       Fax:   (09) 310 5004