Re: Opinion of the value of ordinal methods

Gary wrote:
Norman Cliff, Jeffrey Long and others have been developing statistical
methods based on ordinal methods (mostly Kendall's tau as far as I can
see), and they argue that ordinal questions (and they suggest that
nearly all questions in Psychology, for example, are ordinal) should
be answered with ordinal methods.

See for example

http://www.amazon.com/Ordinal-Methods-Behavioral-Data-Analysis/dp/0805813330/ref=sr_1_3/103-8728016-1667817?ie=UTF8&s=books&qid=1188899201&sr=1-3

http://www.leaonline.com/doi/abs/10.1207/s15327906mbr3103_4?journalCode=mbr

I wondered what people here thought of these methods? Are they worth
the trouble?

Lance

"Are they worth the trouble?" ...you haven't said what you are comparing them to. The alternatives seem to be to do something based on: un-ordered categories; ordered categories; ordered lists of items; partially ordered lists of items; ordered categories with numerical limits; observations of numerical values; observations of numerical values within some of them only known to fall within certain limits; etc.

Basically you should try to make as much use as possible of all the information you have, without introducing information you don't have.

However, perhaps your question should really be "what type of data is it best to collect?" since it is the type of data (characteristics of the data) collected that should determine your choice of method. There have been some recent moves in certain areas to replace questions that effectively say "put these in your order of preference" with "how much would you be prepared to pay to have the use of each of these", thus trying to push things towards using a relative utility scale.

David Jones


.