My comment was purely about grouping. Information is lost by grouping;
that's my
only point.
I don't understand what you understand by quantile here. In particular,
what is a "1-quantile" increase? Perhaps you mean something like
percentile rank,
in effect the inverse of the quantile function.
Nick
[email protected]
Ronan Conroy
On 29 Feb 2008, at 17:59, Nick Cox wrote:
> It's not your question, but most classifications into quantile-based
> groups sound perverse.
> Why throw information away?
Psychometrics
As predictor variables, quantiles have natural advantages over scores
on psychometric instruments such as depression scales or aptitude
scales. These advantages are
1. Easy to understand: the effect size is the effect of a 1-quantile
increase in the predictor
2. Easy to compare: effect sizes for different predictors are
comparable, allowing comparisons to be made
3. The quantiles are based on the actual distribution of the measure
in the population, not on its theoretical score range. The adjust for
the often far-from-nice distributions shown by psychometric measures.
In fact, where the predictor is measured on an unfamiliar and
arbitrary scale, quantiles contain more information (in the
hermeneutic rather than Shannonist sense) than the original scale
values.
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
