Newport iServer MicroServer iTHX-M, Operator'S Manual, Page: 355 / 414

Appendix C | Dynamic Model Mapping Algorithm
347
GENDER CALLS
A gender call is made for the sample by comparing the heterozygous
call ratio for X chromosome SNPs to a threshold value set in the
software (see
Determining Sample Gender, on page 328 ).
ONE-SIDED WILCOXON’S SIGNED RANK TEST
This section describes the technique used by the Dynamic Model
algorithm to calculate rank scores in more detail. The signed rank test
was proposed by Wilcoxon in 1945. The basic procedure is well
known, but there are different techniques used to handle ties and
calculate rank scores.
The signed rank test applies to two paired data sets:
g = (g
1
, g
2
, … , g
n
) and h = (h
1
, h
2
, … , h
n
).
It tests the null hypothesis
H
0
: median(g
i
- h
i
) = 0
versus the alternative hypothesis
H
1
: median(g
i
- h
i
) > 0.
Note that we use the one-sided test here. For the one-sided test, if the
null hypothesis is true, the rank score should be close to 0.5. When the
alternative hypothesis is true, the rank score should be close to 0.
When median(g
i
- h
i
) < 0 is true, the rank score should be close to 1.
As a standard procedure of signed rank test, we first calculate the
differences of all pairs of data:
d
i
= g
i
- h
i
.
In the Dynamic Model algorithm we are using the S values as the
difference values for the paired data set.
If all differences are zero, we output 0.5 as the one-sided rank score. If
some of the differences are zero, we exclude them from further analysis
and use only the nonzero differences for further analysis. For
simplicity, we also denote the remaining nonzero differences as d
i
(i =
1, …, n). We take their absolute values a
i
= |d
i
| ( Table C.1) .
GTYPE_manual.book Page 347 Wednesday, August 2, 2006 11:31 AM