Newport iServer MicroServer iTHX-M Operator's Manual page 349

CALCULATING LOG LIKELIHOOD VALUES (L-VALUES)
The resulting output includes:
- Genotype call
- Confidence score (rank score for genotype call used)
- Rank scores for all genotype calls
In a separate step, the software runs MPAM analysis on a selected set
of SNPs on the Dynamic Model chips for
(see page 345).
The log likelihood value (l-value) provides an initial indication of
which model is the best match to each of the ten individual probe
quartets.
Log likelihood equation is based on observed - estimated values for the
mean.
The log likelihood is calculated as:
4
1 ∑
l ( )
ln = – -- - n
ln
k
2
k = 1
Where for a given probe quartet:
k = index for the probe cell in the quartet (1,2,3,4)
n = number of pixels
k
m = cell intensity
k
v = standard deviation for intensity
k
M = estimated intensity for a given model
k
V = estimated standard deviation for a model
k
The likelihood function for a probe quartet is calculated in the
following steps:
Calculate maximum likelihood estimators from foreground and
1.
background data.
To minimize the estimation error, for each of the four models we
find maximum likelihood estimators for all parameters by using
the assumptions and by differentiating the log likelihood equation
Appendix C | Dynamic Model Mapping Algorithm
(
2
v + m – M
( 2 ) k k k
2πV + ---------------------------------------- -
k
2
V
k
contamination QC check
2
)
341