Figure 2

Parameter estimation by MCMC simulation for a simulated array-CGH data set. (A) The log-ratio vs. probe genomic index plot of a simulated one-CNV array-CGH data set. The data D(M = 500) were generated with θ= {N = 1, (s₁ = 200, w₁ = 50, a₁ = 1.5), σ² = 0.4²}. (B) The logarithm of the posterior probability (calculated up to some multiplicative constant) at consecutive 500 MCMC sampling iterations. In the stationary phase, the posterior probability of the MCMC-sampled parameter values given data D, $p(\widehat{\theta }|D)$, fluctuates closely beneath the maximum value p(θ|D). (C-F) Histograms of the 500 estimates of s₁, w₁, a₁, and σ respectively. (G-J) Traces of the estimates of s₁, w₁, a₁, and σ through the consecutive 500 MCMC sampling iterations.