11.3 Running lm_lods example and sample output

To run the lm_lods example with the phenotypic trait data, open the parameter file `ped73_ph.par' in the `Lodscores/' subdirectory. Then follow the instructions for changing the number of MC iterations. The relevant section of the parameter file is below:

#For actual analyses, recommended number of iterations is #on the order of 10^5 set MC iterations 3000 #1 check progress MC iterations 1000 set global MCMC #TO RUN LM_LODS, comment out the line marked #1 (above), #and uncomment the line marked #2 (below). This effectively #reduces the number of MC iterations #Recall that lm_lods and lm_shnell require less iterations #(order of 10^4) than other programs #set MC iterations 300 #2

Recall that since MCMC is performed at each position, the number of MC iterations needed is less. Once the parameter file has been changed, run the lm_lods example under the subdirectory `Lodscores' by typing:

./lm_lods ped73_ph.par

The most important parts of the output are the LOD scores; these are given at the end of the output for each component (connected pedigree) at each position requested. Since there are 73 individuals in the pedigree, this example takes a while to run.

Below are the LOD scores outputs from this example (some outputs have been omitted to save space):

ESTIMATED LOD SCORES Component 1 The largest eigenvalue : 1.86626 The second largest eigenvalue : 1.57587 Cumulative from left : 2.21620 Cumulative from right : 0.45122 LodScore estimates: Trait pos # position (Haldane cM) or marker male female eigen left right 1 -115.129 -115.129 0.04481 0.00015 -0.34546 2 -80.472 -80.472 0.13091 -0.00410 -0.34971 3 -45.815 -45.815 0.28046 -0.05817 -0.40378 4 -17.834 -17.834 0.43549 -0.18552 -0.53113 5 -5.268 -5.268 0.83469 -0.13949 -0.48510 marker-1 0.000 0.000 NA NA NA 6 3.000 3.000 1.33851 0.00175 -0.34386 7 7.000 7.000 1.68532 0.05630 -0.28931 marker-2 10.000 10.000 NA NA NA 8 13.000 13.000 2.30193 0.17720 -0.16841 9 17.000 17.000 2.58626 0.23244 -0.11317 marker-3 20.000 20.000 NA NA NA 10 23.000 23.000 3.46676 0.62422 0.27861 11 27.000 27.000 3.95936 0.78769 0.44208 marker-4 30.000 30.000 NA NA NA 12 33.000 33.000 5.05419 0.97612 0.63051 13 37.000 37.000 5.42606 1.08006 0.73445 marker-5 40.000 40.000 NA NA NA 14 43.000 43.000 6.14399 1.17323 0.82762 15 47.000 47.000 6.39609 1.23478 0.88917 marker-6 50.000 50.000 NA NA NA 16 53.000 53.000 5.68067 1.06031 0.71470 17 57.000 57.000 5.37402 0.98868 0.64307 marker-7 60.000 60.000 NA NA NA 18 63.000 63.000 4.32190 1.05923 0.71362 19 67.000 67.000 3.91308 1.05841 0.71280 marker-8 70.000 70.000 NA NA NA 20 73.000 73.000 3.35744 1.01417 0.66856 21 77.000 77.000 3.07940 1.04257 0.69696 marker-9 80.000 80.000 NA NA NA 22 83.000 83.000 2.96763 1.33124 0.98563 23 87.000 87.000 2.79748 1.45101 1.10540 marker-10 90.000 90.000 NA NA NA 24 95.268 95.268 1.78970 1.13057 0.78496 25 107.834 107.834 1.13899 0.95320 0.60759 26 135.815 135.815 0.41807 0.59170 0.24609 27 170.472 170.472 0.10067 0.43978 0.09417 28 205.129 205.129 0.01432 0.39120 0.04559

Note that your output may be slightly different due to differences in the random seed. Note also that lm_lods does not estimate LOD scores at marker positions; the markers are included in the list of positions for reference, but the lodscore is given as "NA".

As mentioned in the introduction to this chapter, there are three methods to combine the likelihood ratios (for each test position over the position to the left, and over the position to the right): the eigenvalue method, simple averaging starting from the left, and simple averaging starting from the right.

In theory, the largest real eigenvalue should be equal to 2.0. The eigenvector corresponding to the largest real eigenvalue is given as the LOD scores. However, when the second largest eigenvalue is very close to the largest one, the eigenvector can be very unstable and sometimes gives very bad LOD scores. When that happens, the "left" and "right" method, although simpler, actually perform better.

Ideally the `Cumulative from left' and `Cumulative from right' values at the ends of the chromosome should be log(1) or 0. In practice, the two values rarely equal one and LOD scores differ a lot for these three methods, as can be seen in the output above. This example performed a very short MCMC run. For longer runs, the LOD scores for the three methods may be more consistent with each other.

For more information regarding the MCMC parameters and diagnostic output, See MCMC parameters and options.