A statistical inference from the existing ice thickness data was made. The time series model of ice thickness was established, and the statistical inference from ice thickness was made. Then, the parameter estimation of the inference model by MCMC was conducted on the basis of the Gibbs sampling. At last, error correction of the Markov Chain Monte Carlo(MCMC) model was carried out to solve the error increase problem when the sample data was not enough during the maximum likelihood estimation. A comparison between the Gibbs sampling results by WINBUGS software and the maximum likelihood estimation result was made. By analysis and comparison, it has been proved that the Bayesian method for error correction has a higher accuracy for the estimation of the ice thickness on the cable.