1: \begin{abstract}

2: %% Text of abstract

3: Even though Nearest Neighbor Gaussian Processes (NNGP) 

4: alleviate considerably MCMC implementation of Bayesian space-time models,  they do not solve the convergence problems caused by high model dimension. Frugal alternatives such as response or collapsed algorithms are an answer.

5: An alternative approach is to keep full data augmentation but to try and make it more efficient. Two strategies are presented. \\

6: The first scheme is to pay a particular attention to the seemingly trivial fixed effects of the model. 

7: Empirical exploration shows that re-centering the latent field on the intercept critically improves chain behavior. 

8: Theoretical elements support those observations.

9: Besides the intercept, other fixed effects may have trouble mixing. This problem is addressed through interweaving, a simple method that requires no tuning while remaining affordable thanks to the sparsity of NNGPs. \\

10: The second scheme accelerates the sampling of the random field using Chromatic samplers. This method makes long sequential simulation boil down to group-parallelized or group-vectorized sampling. The attractive possibility to parallelize NNGP density can therefore be carried over to field sampling.   \\

11: A \textsf{R} implementation of the two methods for Gaussian fields is presented in the public repository \\ \url{https://github.com/SebastienCoube/Improving_NNGP_full_augmentation}. An extensive vignette is provided. The presented implementation is run on two synthetic toy examples along with the state of the art package \textsf{spNNGP}. Finally, the methods are applied on a real data set of lead contamination in the United States of America mainland.

12: \end{abstract}

13: