On 01-02-98 William Elliot wrote to John Boone... 
 
        Hello William, 
  
 WE> WE> Now, if you considered the number of arrests to the number of  
 WE> WE> convictions in   
 WE> WE> each group and compared those ratios, a bias would become   
 WE> WE> more apparent as the ratios should be nearly equal.  Yes,   
 WE> WE> this is testing for a different sort of bias, but it has an   
 WE> WE> advantage of being more objective.  
   
 WE> JB> Not necessarily.  It could be coincidence, related to some 
 WE> JB> common factor unrelated to bias, differences of results from  
 WE> JB> equal application of the rules within the system, etc.  
 
 WE> First off the statistics are not complete.  You have to do 
 WE> a chi-squared test to determine the significance of any  
 WE> discrepancies.  If a likely bias is determined it makes no  
 
  The chi-squared test is used to give an idea how likely the 
difference arises from chance alone (from "A Study Guide to 
Epidemiology and Biostatics"). 
   
 WE> difference whether the bias is blatant social bias or from 
 WE> some other systematic cause such as being unable to afford  
 WE> defense.  The determination of the cause is issue for  
 WE> further study, not all necessarily mathematical.   
 
  The word "bias" means (from) Franklin's dictionary 
"predjudice."  For most, the word "predjudice" has 
"racial" tones.  However, as you are probably aware 
from a statistical standpoint (from "A Study Guide 
to Epidemioligy and Biostatics" page 97) "Bias is 
systemic error, resulting in over or underestimation 
of the strength of the association."  From this defintion, 
bias doesn't have necessary -any- "racial" overtunes which 
many often associate with the word "bias." 
  Now, let us look the words: 
 
        Now, if you considered the number of arrests 
        to the number of convictions in each group and 
        compared those ratios, a bias would become more 
        apparent as the ratios should be nearly equal.   
 
  The words "bias would become more apparent as the ratios 
apparent as the ratios should be nearly equal" imply 
differences in ratios would be due to bias, "systemic 
error, resulting in over or underestimation of the 
strength of the association."  When, in fact, WITHOUT 
BIAS, differences in ratios may in fact be due causation, 
etc without "systemic error, resulting in over or underestimation 
of the strength of the association."   
 
Take care, 
John 
 
___ 
 * OFFLINE 1.54 
--- Maximus 3.01
---------------