For identifying QTL-linked markers (the X variables) for a trait (the Y variable), inclusion of (only) one of the co-located or nearby markers looks adequate and that marker should get automatically selected/identified by any step-wise regression approach. Multiple co-located or nearby markers, statistically, will exhibit multiple collinearity, and, therefore, using more than one of these markers may not add much/nothing in terms of genetic variation explained. The link below may help get some idea of multiple collnearity

http://en.wikipedia.org/wiki/Multicollinearity

Go after the haplotipes, the markers are within an LD block, try to define them, if possible.

Thanks for the assistance. So I am gathering that leaving the markers in will not lead to any foreseeable errors in the results.

I don't know what kind of mapping population you are using, but in R/qtl there is a function called jittermap with the following role: Jitter the marker positions in a genetic map so that no two markers are on top of each other. 

see http://solcap.msu.edu/pdf%20files/6PAAextension.pdf 

Those markers should be redundant if having same genetic distances from the top a linkage or chromosome (not from previous marker). Theoretically, using all of them did not provide any additional information over any single marker within the group. A few QTL mapping software did not allow to include redundant markers.

I  wonder why you have so many redundant markers (~ 800 over ~215)? How large is your population size? What types of markers? Perhaps, it's good to double check the methods of genotyping and developing the linkage map.

Your mapping population should be large enough (around 300 non related individuals in a variety panel or the same number in a segregating F1 cross. Large mapping populations are key to avoid redundant markers.

Dear Philip Freda 

I would like to thank you for our question. I think you need to see the segergation of this markers if it is significant or not. Also, I think you could used other QTL software like qgene.

with best regards

Khaled 

Dear Philip

This is a pretty interesting question. I do think that the type of molecular marker that you used matters a lot in your case. I am assuming that you started out with 1000 polymorphic markers.

I think, using R/QTL in your case would be beneficial (from scratch). There are really some great quality control steps (like check for segregation distortion, duplicate genotypes across 2 or more loci, check for switched alleles at a locus etc.). Also, like Dr. Humberto suggested, there is a jitter map function that avoids placing tightly linked markers on the same position. Hope this helps.

Good luck

Raghu

using the redundant markers placed in one position doesn't increase the resolution  of your map or  probability of detection putative qtl in your population. only will be a time consuming analysis for your software. my suggestion is to construct frame work map based on 5  CM  interval between markers then run qtl analysis, after that based on the position of  detected qtl use more marker which you have in that linkage group or position.