Tips

1.  Input

    READ "Input" DOCUMENT CAREFULLY.
 
    The program will stop if any errors are detected in the format of the
    marker data file or in the format of the kinship coefficient file.
    (e.g. missing kinship coefficient values...)  
    So read the "Input" document carefully and make sure the input files
    are in the correct format and have concordant information.

2.  Allele frequency estimation

   Individuals with unknown status (=0) are not taken into account while 
   estimating the allele frequency in the whole sample.
 
   If you only wish to estimate the allele frequencies in a sample using 
   the BLUE estimate proposed by McPeek, Wu and Ober (2004), assign the same
   dummy phenotype to all the indivudals in the sample and consider the all sample
   estimate in the ouput file.

   It may happen in some particular situations that the QL frequency estimates are
   negative. In such cases, the QL computations are skipped. Use naive counting
   estimates and the corrected chi2 statistic.

3. Difference in allele frequency between cases and controls

   The program provides with 3 different allele frequency estimates : 
		- frequencies in the case sample only
	        - frequencies in the control sample only
                - frequencies in the whole sample

  When considering the naive counting estimates of these frequencies, the whole
  sample estimates are a simple combination of the frequencies in the cases and 
  in the controls weighted by their respective sample size.
  When using the corrected chi2 test, the direction of a change in allele frequency
  is directly given by the difference between the case and control samples.

  When considering the QL-estimates, because each observed allele is weighted 
  by some function of its relationship with all the other individuals alleles 
  present in the sample, the link between the whole sample and the case only and
  control only estimates might be complex. Note that the only frequencies used in the
  CC-QLS test are the whole sample estimates. Case only and control only estimates are
  provided for information.
  The sign of the score function should be used to know the direction of a change in 
  allele frequency between cases and control rather than the actual estimates in the 
  case only and control only samples. When the score corresponding to allele j is 
  positivie it means that the frequency of this allele is increased in the cases.


4. Significance

   We have studied the CC_QLS and corrected chi2 statistics in different context and
   found that the chi2 distribution is a good approximation for their null distributions.
   However when the number of alleles of a given type is small in the case or in the 
   control samples, simulations should be preferred to get exact p-values.

5. CC-QLS or corrected chi2 ?

   We have shown in Bourgain et al.(2003) that the CC_QLS test is under certain regularity
   conditions, asymptotically the locally most powerful statistic in a class of statistics 
   that contains the corrected chi2 test.
   However, this does not mean that the CC_QLS will always provide the smallest p-values.
   Further, we cannot rule out that in special cases, our results might not apply. In 
   particular we have not compared the two statistics when the alternative is not local (which 
   might be the case when the difference in allele frequency is large).