A while back I posted a question something like this:  a chi-square analysis 
enables you to conclude whether an observed frequency distribution differs 
significantly from the expected distribution.  How can you determine which of 
the frequency categories is driving the overall departure from expectation?

Several of you wanted to see the responses I got.  Here they are:



1.     Hi, you need to compute the standardized residuals...standardized 
residual (R) tells you where differences lie:



R = O - E/sqrt(E)

where:

O= observed freq

E= expected freq



when R > abs(2), you can conclude it's a major contributor to a significant X2 
value.  (source: Haberman, S. J. 1973. The analysis of residuals in 
cross-classified tables. Biometrics 29:205-220.)



2.        It is possible to calculate the adjusted residuals from a Chi-sq 
test.  An adjusted residual greater than about 2 or 3 in absolute value is an 
indication that the category is pretty different from predicted.


3.        The adjusted residuals are calculated as (obs-exp)/sqrt(exp*(1-prob 
of being in the corresponding row)*(1-prob of being in the corresponding 
column))



4.     I have been using standardized residuals to determine which measurements 
differ from expected, using

+/-2 as a cut-off for significance




George P. Kraemer
  Leff Professor of Environmental Studies and Biology
  Chair, Environmental Studies Program
  Associate Dean, School of Natural and Social Sciences
  Purchase College
  914-251-6640 (o)
 
Website<http://www.purchase.edu/Departments/AcademicPrograms/faculty/GeorgeKraemer/GeorgeKraemer.aspx>

Reply via email to