Academic failures and co-location social networks in campus

EPJ Data Science

Table 1 T-test for the means of changes of degree (or cluster coefficient) for students who failed and who did not

var	Term	End days-start days	Mean of changes		Test statistic (p-value)
var	Term	End days-start days	Students who fail	Students who did not fail	Test statistic (p-value)
Degree	1	126-70	9.56	11.78	−2.25(0.02)
	1	126-91	4.81	6	−2.13(0.03)
	1	126-112	1.12	1.42	−1.31(0.19)
	2	126-70	15.77	19.94	−3.68(0.0)
	2	126-91	8.02	10.5	−4.0(0.0)
	2	126-112	1.21	1.46	−1.26(0.21)
	3	126-70	5.95	8.4	−5.21(0.0)
	3	126-91	3.16	4.88	−5.82(0.0)
	3	126-112	0.91	1.47	−4.31(0.0)
	4	126-70	4.72	5.53	−2.1(0.04)
	4	126-91	2.08	2.59	−2.25(0.02)
	4	126-112	0.27	0.46	−1.98(0.05)
	5	126-70	0.93	1.07	−1.65(0.1)
	5	126-91	0.51	0.59	−1.35(0.18)
	5	126-112	0.12	0.13	−0.39(0.7)
	6	126-70	0.79	0.86	−0.92(0.36)
	6	126-91	0.37	0.39	−0.4(0.69)
	6	126-112	0.1	0.11	−0.27(0.79)
Clustering	1	126-70	0.03	0.03	−0.14(0.89)
	1	126-91	0.02	0.02	−0.49(0.62)
	1	126-112	0.01	0.01	0.21(0.84)
	2	126-70	0.05	0.04	0.54(0.59)
	2	126-91	0.03	0.02	0.69(0.49)
	2	126-112	0.01	0	0.58(0.56)
	3	126-70	0.06	0.02	2.63(0.01)
	3	126-91	0.02	0.02	0.25(0.8)
	3	126-112	0.01	0	0.39(0.69)
	4	126-70	0.05	0.03	1.75(0.08)
	4	126-91	0.02	0.01	1.2(0.23)
	4	126-112	−0.01	0	−1.4(0.16)
	5	126-70	0.02	0.03	−0.76(0.45)
	5	126-91	0.03	0.02	0.42(0.67)
	5	126-112	0.01	0.01	−0.21(0.83)
	6	126-70	0.08	0.05	1.6(0.11)
	6	126-91	0.06	0.04	1.39(0.16)
	6	126-112	0.01	0.01	−0.43(0.67)

We have also built co-location social networks at different time points, including the first 70, 91, 112 days of each term. The changes of degree and cluster coefficients for nodes from these time points to the final snapshot (the end of a semester, 126 days) are then calculated to roughly indicate the evolution of social interactions. This is a two-sided test for the null hypothesis that two independent samples have identical average (expected) values. In the test, we first test using Levene test to see if the populations have identical variances by default. Different t-tests are then applied according to the result of Leven test when populations have either identical variances or different variances.