MODELS

Different parameters for the -M option allow for the analysis of the data assuming different models. Models 1$-$3 were described in Zeng (1993, 1994).

1. Fit all the background markers. This was meant for illustration of the method in the original paper (Zeng, 1993), and should not be used for analysis.

2. Fit all unlinked background markers. This is another illustrative example and also should not be used.

3. Fit only the mean (Lander and Botstein (1989) method)

4. Fit a subset of the other markers, namely those unlinked markers with the highest correlation with the trait on each chromosome. This is an ad hoc model programmed in anticipation of model 6 below.

5. This model uses a pair of markers from each other chromosome and all linked markers that fall outside a window around the flanking markers. This window extends to 10 cM beyond the markers immediately flanking the test position. The window size can be changed with the -w option. This is another ad hoc model programmed in anticipation of model 6 below.

6. This model uses a specified number of markers that fall outside a window around the flanking markers. This window extends to 10 cM beyond the markers immediately flanking the test position. The number of markers are set by the -n option. You need to run SRmapqtl to rank the markers before using model 6. You should use this for composite interval mapping when using markers ranked by SRmapqtl.

7. This model requires that you have already run Zmapqtl and Eqtl. It reads in the estimated QTL from the Eqtl.out file and uses them as virtual markers to control for the genetic background. All identified markers are used that do not fall within the window. This has not been extensively tested: Use it at your own risk.

8. Like model 7, this requires a prior run of Zmapqtl and Eqtl. Instead of using virtual markers, Zmapqtl uses the closest flanking markers to identified QTL. Again, all of these markers outside the window are used. This is good model. You can iterate this process to see if a stable set of cofactors can be identified. See the script below and the Model8.pl(1) man page for more details.

The default is to fit only the mean, that is to use interval mapping.