|
Post by helgebjerck on Mar 28, 2016 13:38:13 GMT
Hi, I'm attempting to use your program to partition the variance in clutch size in sparrows across chromosomes. I have 633 individuals and ~184,000 SNP's, however, I suspect that there is a large amount of relatedness among these individuals. At what level of relatedness does it begin to be problematic? Is there any way to account for this besides simply removing these individuals? If my individuals are too related am I simply out of luck?
Thank you, Helge
|
|
|
Post by helgebjerck on Mar 30, 2016 12:09:27 GMT
As a follow up to the above: When I run my REML with fitting only the single merged GRM I get this: Source Variance SE V(G) 0.158574 0.065085 V(e) 0.663136 0.069070 Vp 0.821710 0.047089 V(G)/Vp 0.192981 0.076910 logL -250.161 logL0 -255.788 LRT 11.255 df 1 Pval 0.0004 n 622
However, when I fit the GRM's for each chromosome simultaneously I get this: Source Variance SE V(G1) 0.000001 0.051652 V(G2) 0.091566 0.068815 V(G3) 0.000001 0.058386 V(G4) 0.000001 0.056710 V(G5) 0.027389 0.055495 V(G6) 0.000001 0.041657 V(G7) 0.000001 0.048622 V(G8) 0.046520 0.051602 V(G9) 0.003825 0.044759 V(G10) 0.000001 0.044938 V(G11) 0.022664 0.040820 V(G12) 0.000001 0.039711 V(G13) 0.007235 0.029510 V(G14) 0.000001 0.040160 V(G15) 0.003668 0.041256 V(G16) 0.002454 0.036175 V(G17) 0.001764 0.035747 V(G18) 0.000001 0.028668 V(G19) 0.083546 0.052472 V(G20) 0.000001 0.027760 V(G21) 0.000001 0.027211 V(G22) 0.000001 0.032618 V(G23) 0.019331 0.048253 V(e) 0.642177 0.089613 Vp 0.952149 0.062708 V(G1)/Vp 0.000001 0.054247 V(G2)/Vp 0.096168 0.071680 V(G3)/Vp 0.000001 0.061320 V(G4)/Vp 0.000001 0.059560 V(G5)/Vp 0.028766 0.058155 V(G6)/Vp 0.000001 0.043751 V(G7)/Vp 0.000001 0.051065 V(G8)/Vp 0.048858 0.054070 V(G9)/Vp 0.004017 0.047008 V(G10)/Vp 0.000001 0.047197 V(G11)/Vp 0.023803 0.042765 V(G12)/Vp 0.000001 0.041707 V(G13)/Vp 0.007598 0.030977 V(G14)/Vp 0.000001 0.042178 V(G15)/Vp 0.003853 0.043323 V(G16)/Vp 0.002577 0.037992 V(G17)/Vp 0.001853 0.037551 V(G18)/Vp 0.000001 0.030109 V(G19)/Vp 0.087745 0.054356 V(G20)/Vp 0.000001 0.029155 V(G21)/Vp 0.000001 0.028579 V(G22)/Vp 0.000001 0.034257 V(G23)/Vp 0.020302 0.050671 logL -247.344 logL0 -247.302 LRT 0.000 df 1 Pval 0.5 n 622
As you can see, the Vp increases substantially from 0.82 to 0.95. This strikes me as odd considering that it's the same phenotype for the same individuals. How is Vp calculated in this instance if not from directly calculating the variance of the phenotype? Is this inflation a result of related individuals?
Thank you, Helge
|
|
|
Post by Jian Yang on Apr 12, 2016 0:11:37 GMT
Hi, I'm attempting to use your program to partition the variance in clutch size in sparrows across chromosomes. I have 633 individuals and ~184,000 SNP's, however, I suspect that there is a large amount of relatedness among these individuals. At what level of relatedness does it begin to be problematic? Is there any way to account for this besides simply removing these individuals? If my individuals are too related am I simply out of luck? Thank you, Helge gcta.freeforums.net/thread/210/use-gcta-greml-family-data
|
|