Subsections


cgtolerance

Fixes the convergence threshold used in EM, the optimization algorithm used for multipoint recombination/breakage/retention estimation.

Synopsis:

The tolerance command is invoked as either:

Description:

The tolerance command sets the convergence threshold used in EM, the optimization algorithm used for multipoint recombination (genetic data), breakage and retention estimation (RH data). The EM algorithm is an iterative loglikelihood algorithm that stops iterating when the loglikelihood increases become too small, below a given threshold.

The first argument is used for final estimation. All the maps presented to the user are converged up to this threshold, initially set to 0.001. The second argument is used for raw estimation inside ordering commands. The loglikelihood is first maximized using this raw threshold (initially set to 0.1) and if it is still far from the loglikelihood of the best know map, then further iteration are abandoned. Otherwise, the iteration are continued until a final level of convergence is reached.

Arguments:

Example:

CG> dsload Data/rh1.cg
{1 haploid RH 13 118 /homes/thomas/carthagene/test/Data/rh1.cg}
CG> sem

Map -1 : log10-likelihood =  -286.91
-------:
 Set : Marker List ...
   1 : G36 MS5 MS6 MS7 MS9 MS8 MS1 G39 MS3 G37 MS15 G40 MS4

CG> dsload Data/rh1.cg
{2 haploid RH 13 118 /homes/thomas/carthagene/test/Data/rh1.cg}
CG> cgtolerance 0.00001 0.01

CG> sem

Map -1 : log10-likelihood =  -286.90
-------:
 Set : Marker List ...
   2 : G36 MS5 MS6 MS7 MS9 MS8 MS1 G39 MS3 G37 MS15 G40 MS4

Thomas Schiex 2009-10-27