Issue link: https://takingitglobal.uberflip.com/i/1542824
R. Adeboye, C. Flewelling,V. Ogbole, E. O'Sullivan 76 ● Age of the session, as Content Providers described how they honed their offerings over time based on practice and feedback. ● Session Subject, given that some subjects may be more readily adapted to a virtual format than others (e.g., with respect to the level of interactivity achievable, technical requirements, etc), vary in the availability of qualified Content Providers, etc. To analyse the teachers' survey questions, binary logistic regression was used. Because responses to these questions were so positive - i.e. largely distributed between 4s (somewhat agree) and 5s (strongly agree), the only feasible option was to code responses as either "no need for improvement" (i.e., the top rating of 5), or "some need for improvement" (i.e., anything lower than the top rating of 5). To build the models, each predictor was assessed individually, then those found to be statistically significant were included in a multivariate model. As there was more variation across response categories in the student survey, ordinal logistic regression was used. Again, to build the models, each predictor was assessed individually, then those found to be statistically significant were included in a multivariate model. Importantly, because the survey responses were not based on a random sample of Connected North users, statistical significance is used only as a guide to the potential importance of predictors rather than as evidence that the relationships would generalize to the wider population of users. Results Details on the models can be found in Annex A. Across the teachers' survey questions tested, multivariate models identified few statistically significant relationships. Where significance emerged, predictors explained little variance, though there were some marked differences in the estimates, particularly across regions. In the student survey, age was consistently predictive across all survey questions. Figure F below - which looks at whether students like coming to school more when there is a Connected North session - is illustrative, noting that these statistics represent the actual measured ratings, and not modelled ratings 39 . 39 Because statistical significance was only used as a guide, the multivariate models ultimately had few predictors, and odds ratios are not easy to interpret at face value, observed values from crosstabulations are provided instead to illustrate the differences.