1: \begin{abstract}

2: Markov Chain Monte Carlo (MCMC) techniques are now widely used for cosmological

3: parameter estimation. Chains are generated to sample the posterior 

4: probability

5: distribution obtained following the Bayesian approach.

6: An important issue is how to optimize the efficiency of such sampling and 

7: how to diagnose whether a finite-length chain has adequately sampled

8: the underlying posterior probability distribution.

9: We show how the power spectrum of a single such

10: finite chain may be used as a convergence diagnostic by means of a fitting

11: function, and discuss strategies for optimizing the distribution for

12: the proposed steps. 

13: The methods developed are applied to current CMB and LSS data interpreted using

14: both a pure adiabatic cosmological model and a mixed adiabatic/isocurvature

15: cosmological model including possible correlations between modes. For 

16: the latter application,

17: because of the increased dimensionality and the presence of degeneracies,

18: the need for tuning MCMC methods for maximum efficiency becomes 

19: particularly acute.

20: 

21: \end{abstract}

22: