Newport iServer MicroServer iTHX-M Operator's Manual page 350

CALCULATING LOG LIKELIHOOD RATIO (S-VALUES)
342
Affymetrix GeneChip ®
with respect to all parameters and solving the corresponding
system of equations. As an example, for model AA, the perfect
match for AA is assumed as foreground, all the other three are
assumed as background and evenly distributed, hence we have:
M = m
1 1
M = M = M =
2 3 4
2
2
V = v
1
1
2 2 2
V = V = V =
4
2
3
Calculate the log likelihood function for each model type using the
2.
estimators and the log likelihood equation.
The log likelihood ratios (LLRs), or S-values, are calculated by
comparing the log likelihood value for a selected model of a probe
quartet to the highest remaining log likelihood value for the quartet:
S
= l - max{l , l , l
N N A B
S = l - max{l , l , l
A A N B
S = l - max{l , l
AB AB N
S = l - max{l , l , l
B B N A
Each of the ten quartets has a set of four model LLRs, or S-values. One
of the models will have a LLR (S-value) greater than zero, and the other
three models will have negative values. The model with the positive
score is the best-fitting model for that quartet
Genotyping Analysis Software User's Guide
⎛ ⎞ ⎛
4 4
⎜ ⎟ ⎜
∑ ∑
n m
⎜ ⎟ ⎜
k k
⎝ ⎠ ⎝
k = 2 k = 2
⎛
4
⎜
∑
[ (
2
n v + M – m
⎜
k k k k
⎝
k =
2
}
AB
}
AB
, l }
A B
}
AB
⎞
⎟
n
⎟
k
⎠
⎞ ⎛ ⎞
4
⎟ ⎜ ⎟
∑
2
) ]
n
⎟ ⎜ ⎟
k
⎠ ⎝ ⎠
k =
2
(Figure C.4).