Page 89 - Applied Probability
P. 89

4. Hypothesis Testing and Categorical Data
                              72
                              1/2. The standardized residual corresponding to c ij is therefore
                                                                    h j
                                                                     2
                                                                      .
                                                       Z ij
                                                                 $
                                                                   h j
                                                                   4
                                                               m
                                                                    2
                                                     2
                              The chi-square statistic χ =    2  =    c ij − Z furnishes an omnibus test for
                                                          i=1  j=1  ij
                              departure from the null hypothesis of Mendelian segregation to affecteds.
                              The maximum standardized residual Z max = max i,j Z ij should be sensitive
                              to preferential transmission of a single allele to affecteds just as in the in
                              situ hybridization and the case-control association problems.
                                The conditional probability space involved in testing the null hypothesis
                              is complicated. Verbally we can describe its sample points as those tables
                              that could have been generated by transmission from the parents with their
                              given genotypes to their affected offspring. The constraints imposed by
                              conditioning on parental genotypes not only fix the margins on the table but
                              also couple the fate of alleles shared by parents. Under the null hypothesis,
                              each relevant transmission event is independent and equally likely to involve
                              either gene of the transmitting parent. Except for biallelic markers [39], it is
                              difficult to compute the exact distribution of either proposed test statistic in
                              this setting. However, we can easily sample from the underlying probability
                              space by randomly selecting for each affected what maternal and paternal
                              genes are transmitted to him. Once these random segregation choices are
                              made, then a new table is constructed by counting the number of alleles of
                              each type transmitted to affecteds. If we let T i be the value of the statistic T
                              for the ith randomly generated table from a sample of n such independent
                              tables, then the p-value of the observed statistic T obs can be approximated
                              by the sample proportion
                                                                   n
                                                                1
                                               Pr(T ≥ T obs )=       1 {T i ≥T  } .
                                               %
                                                                n          obs
                                                                  i=1
                                TABLE 4.3. Transmission/Disequilibrium Test for Costa Rican AT Families
                                    Transmission                    Allele
                                        Pattern      1  3   4   5   7  8  10   11  20   21
                                    Transmitted      3  0   22  0   1  0   0   0    0   2
                                    Not Transmitted  0  4    0  4   3  4   1   1    2   9



                              Example 4.8.1 Ataxia-telangiectasia (AT) in Costa Rica
                                Table 4.3 summarizes marker data on 16 Costa Rican children afflicted
                              with the recessive disease ataxia-telangiectasia (AT). At the chromosome-
                              11 marker D11S1817, 28 of their 32 fully-typed parents are heterozygous.
   84   85   86   87   88   89   90   91   92   93   94