THE TEST OF DOMINANCE AND LINKAGE 



ot Fo and F3. The D and H of the summed variances of backcrosses 

 and of the mean variance of Fg's will, however, differ both from one 

 another and from those of variance of Fo with linkage. If, therefore, 

 we find the best estimates of D and H, E^ and E2, omitting the 



ANTIRRHINUM 



VARIATION IN HEIGHT 



NON-HERITABLE 



-E 

 -H 



FIXABLE HERITABLE -Q 



^m UNFIXABLE HERITABLE 



Vi 



fl 



Vb,+ V82 



"^Fz/n 



Fig. 18. — The components of variation in height of plants of Antirrhinum majus 

 X ghitinosiirn. For each statistic the centre column of the histogram shows the value 

 observed, the left column that expected on the assumption of no linkage, and the 

 right column that expected assuming the possibility of linkage. The compositions 

 of the expected values are shown in terms of D, H, and E. The best estimate of H 

 is negative when linkage is assumed to be absent, so that the contribution of H is 

 overlapped by the contributions of D and E in the left column. When the possibility 

 of linkage is assumed, Vbi + Vb2 and Vps become perfect fits. Wps/ps has no 

 E component. 



Vpo = variance of F2 

 Vbi + Vb2 = summed variances of backcrosses 

 Vf3 — variance of F3 means 

 Vfs = mean variance of Fs families 

 Wf2/f3 = covariance of Fo measurement and F^, mean. 



summed backcross variances and the mean variance of Fg's, these 

 should give a better fit than do the values found for them using 

 all the data, should linkage be interfering with the result. If the fit 

 is not improved there can be no evidence of linkage. 



87 



