Significance of Marker Covariates in CIM

[Jon T]

Dr. Broman,

I am utilizing the CIM() function but am unsure what value to use for  "n.markcovar".

I am having trouble understanding:

1. What marker covariates are and what they do
2. When it is appropriate to use them
3. How you know if you are using too few, too many, or just the right number of marker covariates
4. How the cim() function select markers to be covariates

Here is an example from my data: 

If I set the number of marker covariates to "0", I get the following QTL analysis plot:

If I set the number of marker covariates to "1", I get:

If I set the number of marker covariates to "2", I get:

If you have a moment, I’d appreciate some help understanding what this means.

• Does it mean that by adding two covariates makes the QTL non-significant, the QTL isn't actually there?
• Does it mean that there just aren't any other QTL there besides the large one on the 2nd chromosome?

 

Thank you. I hope you have a pleasant rest of your week! Please let me know if any additional details would aid you. 

[Karl Broman]

The figures didn't show up.

See the end of chapter 7 of the R/qtl book for a discussion of cim(), https://rqtl.org/book. The implementation in R/qtl is rather crude. It fills in any missing data at the markers and then uses forward selection to n.marcovar markers to pick the covariates.

Personally, I prefer stepwiseqtl().

I'm not sure that there is a way to know how many markers to use.

If I saw a single large QTL in scanone(), I might include its genotype as a covariate in a second call to scanone(). Alternatively, you can use makeqtl() and then addqtl(). See Ch 9 of the R/qtl book.

karl

--
You received this message because you are subscribed to the Google Groups "R/qtl discussion" group.
To unsubscribe from this group and stop receiving emails from it, send an email to rqtl-disc+...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/rqtl-disc/07e2f144-4082-4e3f-9b42-475dd2395153%40googlegroups.com.