At the time of writing, the dissertation claimed to introduce a new method for sampling in generalized Gaussian process models, posterior mean centering (PMC). I have now realized that Gamerman (1997) used the same approach, so the claim that the PMC approach was new is incorrect. (Gamerman, D. 1997. Sampling from the posterior distribution in generalized linear mixed models. Statistics and Computing 7:57-68.) On p. 119, in (4.1), the 2*pi in the denominator should be raised to the power m/2 rather than to the power 1/2 as it is now. On p. 120, (4.3) and the preceeding equation are bogus. The inner integral in (4.2) should mirror the expression in (4.1). For small datasets of test values, the outer integral would need to be approximated by a Monte Carlo estimate by averaging over multiple test datasets. However for a large dataset, the outer integral should be approximately equal to the value inside the integral for a single dataset, since we are already averaging over all the values in the test dataset. Also, in (4.2), the first term (the fraction) should either be multiplied by m (the length of the test set) or the integral should be divided by m, so that both terms are on the same scale. Note that lpd.simdata.q, in which the calculations were actually done, appears to be correct, so the results in the thesis should be correct. In the spatial model, for the prior distribution for kappa_Y, kappa_alpha, kappa_beta, and kappa_eta, I used a U(-5.4,1.2) distribution on the log of the paraemters, not a N(-5.4,1.2^2) as indicated in Appendix A of the thesis. On p. 201, in chapter 5, I incorrectly state the null hypothesis. The null should of course be that H_{0,i}: \beta_i \tilde{\beta_i} > 0 , while the alternative is that the true slope is of the opposite sign from the estimated slope. on pp. 62, the matrix \Gamma_1 should be 3X3, with the 2X2 matrix given there as the lower 2X2 corner of the 3X3 matrix and with the 1,1 position a 1, the 1,2, 1,3, 2,1, and 3,1 positions all zero. On p. 31, eqn. 2.7, neither the "2" inside the square root in the argument that is raised to the 'nu' power nor the "2" inside the square root in the Bessel function argument should be included in the square root. These arguments should be 2*sqrt(nu*Q_ij)