See Liu, Statistical Genomics: Linkage, Mapping, and QTL Analysis. pgs 99-100.
"The ratio between the likelihood of Q taking a value QA, which maximizes the likelihood, e.g. a maximum likelihood estimator and the likelihood of Q under the null hypothesis QN, e.g. QN = 0.5 is: L(QA/X) "
L(QN/X)
Likelihood ratio test statistic = G
| = -2ln | [ | l(QA|x) | ] | |
| l(QN|x) |
G has an appropriate X2 distribution.