Skip to Main Content
Table 3. 
Model misspecification is reduced by constraining substitutions to the local disciplinary networks. We computed Kullback-Leibler (K-L) divergences between empirical and simulated journal pair frequencies using two different background networks (local versus global) for each disciplinary network (applied physics, immunology, and metabolism) for the years 1985, 1995, and 2005. K-L divergence was calculated using the R seewave package (Sueur et al., 2008). For every disciplinary network, there is a smaller K-L divergence between simulated and observed data when using the local network (i.e., the disciplinary network) as compared to the global network (all of WoS). Put differently, model misspecification is reduced in the constrained model compared to the unconstrained model.
Disciplinary networkYearBackground networkK-L divergenceRatio
Applied physics 1985 local 1.21  
1985 global 2.37 1.96 
1995 local 0.86  
1995 global 2.37 2.77 
2005 local 0.95  
2005 global 2.35 2.47 
Immunology 1985 local 0.75  
1985 global 1.68 2.24 
1995 local 0.78  
1995 global 1.70 2.19 
2005 local 0.73  
2005 global 1.92 2.63 
Metabolism 1985 local 1.11  
1985 global 2.24 2.02 
1995 local 1.07  
1995 global 2.33 2.17 
2005 local 1.19  
2005 global 2.60 2.18 
Disciplinary networkYearBackground networkK-L divergenceRatio
Applied physics 1985 local 1.21  
1985 global 2.37 1.96 
1995 local 0.86  
1995 global 2.37 2.77 
2005 local 0.95  
2005 global 2.35 2.47 
Immunology 1985 local 0.75  
1985 global 1.68 2.24 
1995 local 0.78  
1995 global 1.70 2.19 
2005 local 0.73  
2005 global 1.92 2.63 
Metabolism 1985 local 1.11  
1985 global 2.24 2.02 
1995 local 1.07  
1995 global 2.33 2.17 
2005 local 1.19  
2005 global 2.60 2.18 
Close Modal

or Create an Account

Close Modal
Close Modal