Figure 3
From: Multiple trait multiple interval mapping of quantitative trait loci from inbred line crosses
Comparison of performances between ECM and GEM-NR algorithms. Comparison of performances between ECM and GEM-NR algorithms in terms of number of iterations required to the convergence of the likelihood function. Both algorithms were applied to an MTMIM model of traits PC1 and ADJPC1 of the BM data. The algorithms were said to have converged whenever the difference between the natural logarithm of the likelihood function of two consecutive iterations was smaller than or equal to 10−4. (A) shows the values of the natural logarithm of the likelihood function at each iteration [log e (L k )] until convergence was reached. The GEM-NR algorithm began with 5 iterations of ECM algorithm. Therefore, the first 5 iterations produced identical values in the likelihood function of both algorithms, and because of that we omitted the first 4 iterations. (B) shows the difference between the natural logarithm of the likelihood function of two consecutive iterations until convergence was reached. In (B), the y-axis was rescaled via logarithm of base ten to improve graphical resolution.