Can changing the step size in conditional genotype probabilities lead to different results?

[Lena Flörl]

Hi Forum,

I am performing QTL mapping with a lot of phenotypes on a 4-way cross, with HK.

First I ran the scan with step = 10 but then - since a "smaller step size is always better" - reduced it to step = 1.

But this drastically altered the results! 

E.g for some phenotypes I got 54 sign. QTLs before and now only 5, wheres for others I had 10 QTLs with step=10 while step=1 yielded 56.

How can this be? From my understanding this just changes the density of the grid on which the genotype probabilities are calculated - but how does that trickle down to such a dramatic change of results? 

I will still use the smaller step, if this is the proper way of doing it - but I'm just curious what causes the shift!

Thanks a lot!

Kind regards, 

Lena 

[Karl Broman]

I can see that it's possible but it is unusual. It suggests a great deal of missing genotype information. It's hard to say much without seeing the details.

Did you determine the significance of QTL using a permutation test? 

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/acf9169b-67c7-41eb-89b1-dc94559b0e2en%40googlegroups.com.

[Lena Flörl]

Hi Karl, 

thanks a lot for the super quick reply!

Yes, I do permutation testing (1000 permutations) for each individual phenotype and have app 2600 markers covering a 500 Mb genome (19 chromosomes). I double checked but cannot see what else changed in my script - except for the step parameters. 

Thank you again! 

Kind regards, 

Lena