Re: Gamma Correlation pro's & con's

Rich Ulrich wrote:

> I think the polychoric is used more for "estimates,"
> with normality as a strong assumption, than for tests.

I think what Rich refers to here (please correct me if incorrect) is
the difference between the use of a statistic (1) to test *whether*
there is statistically significant association (i.e., to provide a
p-value) versus (2) to estimate the *amount* of association.

If the original poster only needs a statistic for (1), that might be an
important detail. For example, even though, say, Spearman's rho is
affected by ties, do they affect its use as an "estimate" more than its
use as a "test"?; perhaps many ties might make the significance test
more conservative--if so, if one gets a statistically significant
value, there is no problem;one can still conclude a significant
association.
--
John Uebersax PhD

.

Relevant Pages

  • Re: Gamma Correlation pros & cons
    ... Rich Ulrich wrote: ... > with normality as a strong assumption, ... affected by ties, do they affect its use as an "estimate" more than its ... if one gets a statistically significant ...
    (sci.stat.edu)
  • Re: What test is suitable for multiple groups?
    ... phenomenon is statistically significant. ... 1000 B's would be informative as shaping the profile of Jp. ... There's a subtle question of the nature of the inference ... Maybe as Rich Ulrich points out, ...
    (sci.stat.math)
  • Re: What test is suitable for multiple groups?
    ... and perform Chi-square test then? ... phenomenon is statistically significant. ... Maybe as Rich Ulrich points out, ...
    (sci.stat.math)
Loading