Figure 3.

Maximum likelihood estimation and search. (A) Log likelihoods over different values of <a onClick="popup('http://genomebiology.com/2013/14/8/R90/mathml/M10','MathML',630,470);return false;" target="_blank" href="http://genomebiology.com/2013/14/8/R90/mathml/M10">View MathML</a> (the es-timate of α) and <a onClick="popup('http://genomebiology.com/2013/14/8/R90/mathml/M11','MathML',630,470);return false;" target="_blank" href="http://genomebiology.com/2013/14/8/R90/mathml/M11">View MathML</a>AA,AB (the estimate of GAA,BB, the probability that the control genotype is AA and the disease genotype is AB). The surface graph shows that likelihoods are maximized around the true α values (10% (left), 50% (middle), 90% (right)). (B) Search paths using the feasible direction method on contour maps. The method efficiently finds the optimum points (green circles) in only a few search steps. Searches from different starting points (blue circles) finally converge.

Kim et al. Genome Biology 2013 14:R90   doi:10.1186/gb-2013-14-8-r90
Download authors' original image