EPJ Data Science

Table 2 Regression results for the tie persistence using generalized linear models for the two prediction models

From: Temporal patterns behind the strength of persistent ties

Feature	Model 1			Model 2
Feature	Full	Simplified	Simplified \(\boldsymbol{'}\)	Full	Simplified
\(w_{ij}\)	1.491^∗∗∗	1.180^∗∗∗		2.096^∗∗∗	2.229^∗∗∗
\(w_{ij}\)	(0.021)	(0.015)		(0.060)	(0.057)
\(d_{ij}\)	0.105^∗∗∗			0.109^∗∗
\(d_{ij}\)	(0.012)			(0.034)
\(r_{ij}\)	−0.094			−0.044^∗∗
\(r_{ij}\)	(0.012)			(0.032)
\(o_{ij}\)	0.241^∗∗∗			0.286^∗∗∗
\(o_{ij}\)	(0.017)			(0.046)
\(k_{ij}\)	0.039^∗∗			−0.026
\(k_{ij}\)	(0.012)			(0.033)
\(\mu _{ij}^{\mathrm{int}}\)	0.084^∗∗∗			0.151^∗∗
\(\mu _{ij}^{\mathrm{int}}\)	(0.012)			(0.032)
\(\mathit{age}_{ij}\)	0.001			−0.021
\(\mathit{age}_{ij}\)	(0.012)			(0.034)
\(\mathit{gender}_{ij}\)	0.079^∗∗∗			0.035^∗
\(\mathit{gender}_{ij}\)	(0.012)			(0.033)
\(\hat{f}_{ij}\)	−1.102^∗∗∗	−0.660^∗∗∗	−0.611^∗∗∗
\(\hat{f}_{ij}\)	(0.016)	(0.008)	(0.007)
\(\mathit{cv}_{ij}\)	−0.362^∗∗∗			−0.653^∗∗∗	−0.759^∗∗∗
\(\mathit{cv}_{ij}\)	(0.014)			(0.037)	(0.034)
\(\mu _{ij}^{\mathrm{chats}}\)	0.029^∗			0.036
\(\mu _{ij}^{\mathrm{chats}}\)	(0.014)			(0.038)
\(a_{ij}\)	−0.310^∗∗∗			−0.316^∗∗∗
\(a_{ij}\)	(0.013)			(0.036)
Constant	0.681^∗∗∗	−2.364^∗∗∗	1.053^∗∗∗	0.779^∗∗∗	0.748^∗∗∗
Constant	(0.014)	(0.042)	(0.014)	(0.039)	(0.034)
Number of points	45,444	45,444	45,444	6684	8268
AUC	0.864	0.847	0.755	0.875	0.866

Performance	Model 1			Model 2
Performance	Full	Simplified	Simplified \(\boldsymbol{'}\)	Full	Simplified
Accuracy	0.802	0.766	0.721	0.803	0.796
Sensitivity	0.828	0.777	0.652	0.815	0.808
Specificity	0.767	0.753	0.781	0.787	0.760

Coefficients are shown with uncertainties (standard errors) in parentheses. Model Full include all the features described in the text, while model Simplified/Simplified′ only includes the most important two/one feature(s). Note: ^∗ p<0.1; ^∗∗ p<0.05; ^∗∗∗ p<0.01.

Back to article page