Re: Ordinal variable considered as quantitative ?

Loïc BRANCHERIAU wrote:
Dear all,

I am new in the forum and my question is about ordinal variable.
The details of the experiment are :
- A xylophone maker was asked to classify 60 sounds (60 xylophone bars made
with different species of wood). So the sounds were ranked from 1 (the best
quality) to 60 (the worst).
- The wooden bars were then characterized by acoustic parameters (duration,
brightness, temporal damping, spectral center of gravity...) and by material
parameters (young's modulus, density...). All these parameters are
quantitative variables.

The main objectives of the study are:
- Which parameter(s) explain the classification?
- does exist a combination of parameters which can estimate the
classification?

To reach the objectives, can I use the multivariate regression analysis (PCR
and MLR) considering the ordinal variable Classification as quantitative? Do
exist other multivariate analysis methods to be used with an ordinal
variable?

Many thanks,

Loïc

How many acoustic and material properties were measured? With only 60
cases, it may matter. Also, it might be a good idea to avoid the term
"parameter" in this discussion, because the way you have used it could
be confused with its statistical meaning.

Rankings of this sort tend to be more reliable at the good end than at
the bad. Is the goal more to distinguish the excellent from the merely
good than to discriminate over the full range?

In any case, you should start by looking at scatter plots of quality
with each property. Is there any reason to expect quality to be a
non-monotone or interactive function of the properties?

.

Relevant Pages