Variance explained

[Othmane Lamoumni]

Hello everyone,

I would like a little help to understand the %var information obtained through fitqtl(dropone = T).

Does it mean:

- The % var explained by a QTL in a single model including that QTL alone.

- The % var explained by a QTL in the context of the established multiple QTL model, which in this case, the sum of %var explained I assume should be less than 100%?

                        LOD            %var   
3@270.0      20.617          8.005 
4@94.2        27.450         12.806 
8@202.0      40.786         27.817 
10@249.0    24.093         10.261 
11@257.0    18.538           6.816             
12@244.5    10.345           3.092          
13@265.5    30.401         15.389             
21@123.5    18.478           6.784            
21@133.5    27.992          13.255         
22@163.7    16.870            5.941 

Othmane

[Karl Broman]

Those percent variance explained are derived simply as a transformation of the LOD scores from the drop-one-QTL analysis, and so they can’t be fully trusted. They should be approximately correct if the loci were all completely unlinked and additive; they’re especially untrustworthy in the case of linked QTL.

karl

[Othmane Lamoumni]

Indeed, I've come to check the function description. It states these estimations can be completely wrong in the case of linked QTL, as alternatives aren't available yet.