value with the specified range is equally likely to be correct. To evaluate the influence of the prior distribution on the posterior pdf, the algorithm was repeated using normal marginal pdfs as the prior for the force coefficients. The marginal prior pdfs were selected as:
Kt (N/mm2) = N(2500, 300)
Kn (N/mm2) = N(1200, 300)
Kte (N/mm) = N(100, 33)
Kne (N/mm) = N(100, 33).
Table 6 lists the mean and standard deviation for each of the four force coefficient posterior marginal pdfs. Figure 7 provides a comparison between the prior marginal pdfs and posterior sample histograms of the force coefficients. The percent errors in Table 6 (normal prior) are larger than those in Table 3 (uniform prior). The posterior distribution is clearly sensitive to the choice of the prior.
Table 3Comparison of the force coefficient distributions from MCMC to the true values using a uniform prior
Table 6Comparison of the force coefficient distributions from MCMC to the true values using a normal prior