13.4 lm_map statements

use MCEM and SA for maximization

If the statement `use MCEM only for maximization' is replaced by this statement, lm_map will attempt to refine its MCEM based estimate of the MLE by performing additional SA steps.

set SA curvature iterations I

An estimate of the curvature of the likelihood is needed to initiate the SA algorithm. This statement tells lm_map to use at least I MCMC realizations to estimate the curvature of the likelihood. The curvature is only estimated once.

set SA ascent iterations I

lm_map will not initiate the SA algorithm with a step that decreases the likelihood. This statement tells lm_map to use at least I MCMC realizations to determine whether a proposed first step increases the likelihood.

set SA gradient iterations I

If SA is initiated, this tells lm_map to use at least I MCMC realizations to estimate the gradient of the likelihood. An estimate of the gradient is needed for each SA step.

set SA convergence R

The SA algorithm is terminated, if all recombination fraction updates are within R of their previous values. In addition, the maximum possible runtime for the SA algorithm is proportional to the total runtime of the MCEM algorithm.

set LRT stat iterations I

This statement tells lm_map to use at least I MCMC realizations to estimate the LRT statistics. If only one option is used in `output maps gender ...', then the estimated LRT statistic compares the MLE to the initial map. Otherwise, two LRT statistics are estimated. The first compares the MLE of the sex-averaged map to the initial sex-averaged map, while the second compares the MLE of the sex-specific map to the MLE of sex-averaged map.

compute estimates I times

This statement tells lm_map to conduct its entire analysis I times, and to report the map with the highest likelihood. While this statement offers some protection against convergence to local modes, the default value is 1.