Table 2 

Summary of generalized linear model results. Each row corresponds to a single model using multiple explanatory variables. Significant values are indicated in bold. Quasibinomial and negativebinomial error distributions were used and results are given after calculating typeII analysisofvariance using a F test (Quasibinomial family) or a likelihood ratio test (negativebinomial family). 

Explanatory variables 



Upstream parameters 
AFLP parameters 
AFLP diagnostics 





Species 
No. of sb^{a} 
CG cont 
No. of peaks per profile 
Peak length 
Homoplasy per peak 

Response variables 
Distribution 
(Df = 2) 
(Df = 1) 
(Df = 2) 
(Df = 1) 
(Df = 1) 
(Df = 1) 


In silico analyses 

Peak length distribution^{b} 
Quasibinomial 
F_{1,1349 }= 2.09e12, P = 1 
F_{1,1349 }= 967.20, P < 0.001 

No. of peaks per profile 
Negative binomial 
χ^{2 }= 1574.87, P < 0.001 
χ^{2 }= 1869.47, P < 0.001 
χ^{2 }= 275.24, P < 0.001 

Homoplasy rate H 
Quasibinomial 
F_{2,278 }= 0.98, P = 0.38 
F_{2,278 }= 0.72, P = 0.49 
F_{1,278 }= 576.26, P < 0.001 

No. of CF^{c} 
Negative binomial 
χ^{2 }= 19.90, P < 0.001 
χ^{2 }= 2.99, P = 0.22 
χ^{2 }= 25.32, P < 0.001 
χ^{2 }= 37.92, P < 0.001 

No. of CF^{c }in peaks with more than 10 fragments 
Negative binomial 
χ^{2 }= 6.96, P = 0.03 
χ^{2 }= 0.90, P = 0.64 
χ^{2 }= 0.87, P = 0.35 
χ^{2 }= 2.74, P = 0.10 

Empirical analyses 

Fluorescence intensity 
Negative binomial 
χ^{2 }= 22.33, P < 0.001 
χ^{2 }= 2.39, P = 0.12 



^{a }Total number of selective bases added for the 2 primers. ^{b }Peak length distribution is expressed in relative frequencies. ^{c }"CF" corresponds to comigrating fragments within a peak. 

Paris et al. BMC Genomics 2010 11:287 doi:10.1186/1471216411287 