### USING MRQAP TO ANALYSE THE DEVELOPMENT OF MATHEMATICS PRE-SERVICE TRAINEES’ COMMUNICATION NETWORKS

1. USING MRQAP TO ANALYSE THE DEVELOPMENT OF MATHEMATICS PRE- SERVICE TRAINEES’COMMUNICATION NETWORKS Christian Bokhove and Chris Downey University of Southampton Jasperina Brouwer University of Groningen
2. Context Teacher training in UK • PGCE University Led (UL) • School Direct (SD) → Newly Qualified Teacher Secondary Maths and Science • cohort size (~35) • “sink or swim” • longevity of course Social (support)Networks play a role
3. Social Network Analysis (SNA) as method
4. Dynamic SNA From a previous project: networks change over time.
5. Data collection • Longitudinal - 4 ‘waves’ of data collection (per 2 months) • Directed network question: “During the last month, to whom have you turned for support?” • Both instrumental and affective aspects of support • Online questionnaire instrument Network Related factors Peer (whole) External (ego) Trust Network intent Support views Self- efficacy 1     2      3       4       See Bokhove & Downey (2018) for more info on measures
6. Longitudinal development • Visual representation of the development of mathematics trainees’ communication networks. • Red nodes trainees, arrows indicate communication • For every trainee, every red node, the attributes gender, program, trust and self-efficacy were collected • The figure shows a development from a tight knit community to a situation where some trainees go their own way.
7. MRQAP • Studies have used multiple (logistic) regression or Quadratic Assignment Procedures (QAP) for a bivariate association between two networks. • But longitudinal data violate independence. • Multiple regression quadratic assignment procedures (MRQAP) • In our case we would be able to test whether gender, program, trust or self-efficacy predict the development of communication networks. • The interpretation of the standardised betas is similar to OLS regression analyses. • We use R statistical software with the asnipe package.
8. Results The program (UL or SD) is a negative predictor in all final models. Trust only is a positive predictor from timepoints 3 to 4, perhaps indicating that by that time, towards the end of the year of training, it is trusted peers that are mainly sought. The table shows that the configuration of previous networks predict future networks.
9. Other methods – other SNA • Teacher student interaction • Whole class, directionality • Student interactions • Groups and cliques • Individual behaviour • Help seeking • Disturbances • Central students • Perhaps, patterns over classes, schools, countries (analogue TIMSS video study) • Longitudinal: STERGM, RSIENA
10. Conclusions • Social phenomena can be modelled with SNA. • Networks change; how can we describe this change? • Visualisations. • Descriptive statistics and inferential statistics. • MRQAP is one example of this. • Working on a comparison article of popular methods (Bokhove, Brouwer, & Downey, in preparation). • These methods can be complemented by rich qualitative data (and vice versa).