lod.int plot from scantwo

[mas]

Sorry if the following has come up before but I'm new to the group.

I have a question about plotting scantwo output. The upper default is
apparently “lod.int” but the plots (based on lod score, positions,
etc.) don’t match the “lod.int” results in the summary of the scantwo
output. Even when I assign lower and/or upper=”lod.int” or “lod.i”,
the plot doesn’t seem to match any of the lod results (meaning,
lod.fv1, lod.add, etc.), so I'm not sure what it's plotting. But when
I assign lower=”fv1”, e.g., the plot matches the summary results
perfectly. Sorry to bother but this is after checking that I’m
plotting and summarizing the same output, have confirmed the issue by
running on separate data files, etc. I tend to believe that the
summary is correct, as what is listed under lod.int matches lod.full –
lod.add. So, how to get lod.int to plot?

Any feedback would be most appreciated. Thanks in advance.

[Karl Broman]

The interaction lod score in the plot of scantwo results is pointwise: it's the difference between the full lod score at a position and the additive lod score at that position.

The interaction lod score that you see in the summary of scantwo results is, rather different. For a given pair of chromosomes, we find the positions that give the biggest full lod score and then find the positions that give the biggest additive lod score. The interaction lod score shown is the difference between these two.

If you want a summary of the sort of pointwise interaction lod scores as appear in the scantwo plot, you need to use the argument what="int" when you call summary. That will then find, for each pair of chromosomes, the positions with the largest pointwise interaction lod scores, and then give the values for all of the lod scores for that pair of QTL positions.

Ponder these:

data(hyper)
hyper <- subset(hyper, chr=1:4) # to save time
hyper <- calc.genoprob(hyper)
out <- scantwo(hyper, method="hk")

summary(out)
summary(out, what="int")

regards,
karl

> --
> You received this message because you are subscribed to the Google Groups "R/qtl discussion" group.
> To post to this group, send email to rqtl...@googlegroups.com.
> To unsubscribe from this group, send email to rqtl-disc+...@googlegroups.com.
> For more options, visit this group at http://groups.google.com/group/rqtl-disc?hl=en.
>

[mas]

Thanks Karl, that's just what I was after.

kind regards,

V

[mas]

As a follow up, when one runs a scantwo permutation test, you again
get multiple outputs (one for full, add, int, etc.) - does the int
output here reflect the pointwise comparison or the int results based
on the biggest full lod score, etc.? If the latter, then is there a
way to do a permutation test for the pointwise scantwo data?

Many thanks in advance.

mas

[Karl Broman]

Yes, in the permutations the lod.int is for a comparison of the best full model to the best additive model. If you want to look at pointwise LOD scores you'll have to write code to do the permutations directly.

karl

[mas]

Got it. Thanks again!

On Apr 3, 2:31 pm, Karl Broman <kbro...@gmail.com> wrote:
> Yes, in the permutations the lod.int is for a comparison of the best full model to the best additive model. If you want to look at pointwise LOD scores you'll have to write code to do the permutations directly.
>
> karl
>