Can I use Chauvenet's criterion on set of observations where a normal distribution cannot be assumed?
$\begingroup$
$\endgroup$
4
-
3$\begingroup$ Surely, you are aware that robust statistics and outlier detection has improved a lot in the last 150 years. Is there any reason you are using this test? $\endgroup$– user603Commented May 14, 2014 at 17:08
-
$\begingroup$ @user603, thanks soo much! I am reading "An Introduction to Error Analysis: The Study of Uncertainties in Physical Measurements" of John R. Taylor (chapter 6) and I thinked Chauvenet's criterion can be use only on set of observations where a normal distribution can be assumed! $\endgroup$– mleCommented May 14, 2014 at 17:34
-
2$\begingroup$ You might want to ditch Chauvenet altogether. Look perhaps at this answer for a way to identify univariate outliers without assuming symmetry. $\endgroup$– user603Commented May 14, 2014 at 17:54
-
$\begingroup$ @user603, ah.. more interesting..!! thanks a lot for hint! ;) $\endgroup$– mleCommented May 14, 2014 at 18:04
Add a comment
|