9.6.8 Autozyg MCMC parameters and options

These statements which set the parameters for the MCMC algorithms, apply to both Autozyg programs, and to the location LOD score programs, lm_lods, lm_markers, lm_bayes and lm_schnell unless otherwise noted.

use (locus-by-locus sampling | sequential imputation) for setup

There are two setup methods available to find a starting configuration for the meiosis indicators prior to the MCMC: using sequential imputation (with the trait treated as unlinked), or using locus-by-locus sampling (by assuming all markers and trait are unlinked). Sequential imputation is the default method.

use I sequential imputation realizations for setup

When sequential imputation is selected above, this statement specifies the number of sequential imputation samples from which the starting configuration of meiosis indicators is to be selected. The default is 20 iterations.

set MC iterations I

Required. It specifies the total number of main L- and M-sampling iterations. There is no default number of MCMC iterations; the total number of "main" L- and M- sampling iterations must be specified for all Autozyg programs. The total MCMC run length is the sum of the number of burn-in iterations specified by the `set burn-in iterations' statement and the number of main iterations specified in `set MC iterations'.

set burn-in iterations I

Burn-in iterations are performed initially, with the trait locus (if any) unlinked to the marker map. The default number of burn-in iterations is specific to each program.

sample by (scan | step)

By default (sample by scan), all loci (L-sampler) or all meioses (M-sampler) are updated successively in an order determined by random permutation. When sampling by step, a single locus (L-sampler) or single meiosis (M-sampler) is randomly selected for updating. lm_bayes presently samples by scan only.

set L-sampler probability X

The L-sampler probability, between 0.0 and 1.0, specifies the probability in each MCMC iteration, of locus-sampling rather than meiosis-sampling. The default is 0.0, that is, to use M-sampler only.

compute scores every I iterations

The default is to score recombinations, total log-probabilities or the Rao-Blackwellized estimator every MCMC iteration. This statement specifies the frequency with which to compute the contributions to the ibd scores or the location LOD scores.

check progress I MC iterations

Use this statement to monitor the progress of the program as it is running. It will print out the iteration number every Ith iteration.