Comparing summaries of imputations

[CG Oakley]

I am wondering if anyone can offer advice (or code) on summarizing
genotypes across imputations?

I would like to determine how much the results of an analysis might
change when run with different numbers of imputations, but it is much
too slow to wait for the full analyses.

Instead I thought I could look at the average (across imputations)
genotypes for each locus for say 32 imputations and again for say 512
imputations. If there is no difference, and this is repeatable, then
it seems safe to use the lower number of imputations for all future
analyses. Does this seem reasonable?

I know enough to save the imputations to file (separately by
chromosome), but am having trouble coming up with a functional bit of
code to summarize the file. Part of my problem is that actual marker
locations and imputed markers have different column name pattern, and
my grep-ing skills are not so great.

Any advice appreciated.

Thanks,

Chris

[Karl Broman]

You might look at a plot of the missing information ( with plot.info), but I think it would be best to look at results with independent sets of imputations. If there are no important differences with 32 imputations, it seems safe to use that number.

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.
>