Closed kieranatkins closed 3 months ago
It's pretty hard to avoid this sort of problem, which generally occurs in regions where one of the 19 haplotypes is likely absent.
# which position and haplotype has min estimated effect?
which.min(apply(effects, 1, min)) # position 208
which.min(apply(effects, 2, min)) # haplotype 12 (Po)
# maximum probability for that haplotype at that position
max(a_probs[[1]][208,12,]) # 0.02
I'm personally not too enthusiastic about the use of scan1coef()
. I prefer to focus on the estimated effects at an inferred QTL, rather than looking at the estimated effects across a whole chromosome.
Ah that makes a lot of sense when considering the probabilities at that position. Thanks for clearing that up. I may switch to looking at effects at the QTL, thanks again.
Hi all,
I am running a QTL analysis on some custom phenotype data on the Arabidopsis Thaliana MAGIC kover2009 data, I have taken the genotype data from the atMAGIC (https://github.com/tavareshugo/atMAGIC) library, however a few of my phenotypes return extreme values for the scan1coef (way beyond the max and min in that actual data series)
I have written a small script to reproduce the error, along with a link to a csv file containing the data. Please let me know if this is expected behaviour, or if I have made a mistake in my script.
Thanks.
Code to reproduce error:
Which returns
Much higher values than I would expect.
Data: reduced_pheno.csv