Analytic and holistic marking are typically researched as opposites, generating a mixed and inconclusive evidence base. Holistic marking is low on content validity but efficient. Analytic approaches are praised for transparency and detailed feedback. Capturing complex criteria interactions, when deciding marks, is claimed to be better suited to holistic approaches whilst analytic rules are thought to be limited. Both guidance and evidence in this area remain limited to date. Drawing from the known complementary strengths of these approaches, a university department enhanced its customary holistic marking practices by introducing analytic rubrics for feedback and as ancillary during marking. The customary holistic approach to deciding marks was retained in the absence of a clear rationale from the literature. Exploring the relationship between the analytic criteria and holistic marks became the focus of an exploratory study during a trial year that would use two perspectives. Following guidance from the literature, practitioners formulated analytic rules drawing on their understanding of the role of criteria, to explain output marks by allocating weightings. Secondly, data derived throughout the year consisting of holistic marks and analytic judgements (criteria) data were analyzed using machine learning techniques (random forests). This study reports on data from essay-based questions (exams) for years 2 and 3 of study (n = 3,436). Random forests provided a ranking of the variable importance of criteria relative to holistic marks, which was used to create criterion weightings (data-derived). Moreover, illustrative decision trees provide insights into non-linear roles of criteria for different levels of achievement. Criterion weightings, expected by practitioners and data-derived (from holistic marks), reveal contrasts in the ranking of top criteria within and across years. Our exploratory study confirms that holistic and analytic approaches, combined, offer promising and productive ways forward both in research and practice to gain insight into the nature of overall marks and relations with criteria. Rather than opposites, these approaches offer complementary insights to help substantiate claims made in favor of holistic marking. Our findings show that analytic may offer insights into the extent to which holistic marking really aligns with assumptions made. Limitations and further investigations are discussed.

Despite the success of plastic bag charges in the UK, there are still around a billion single-use plastic bags bought each year in England alone, and the government have made plans to increase the levy from 5 to 10 pence. Previous research has identified motivations for bringing personal bags to a supermarket, but little is known about the individuals who are continuing to frequently purchase single-use plastic bags after the levy. In this study, over a million loyalty card transaction records from a high-street health and beauty retailer were harnessed to study 12,968 individuals' bag buying behaviour (analysed using descriptive statistics). Statistical regional differences in plastic bag buying throughout the UK occurred. From the transaction data 2,326 frequent single-use plastic bag buyers were identified and matched randomly to infrequent buyers, creating a balanced sub-sample which was used for predictive modelling (N =4,652). For each individual in the modelling sample, their transaction data was matched to questionnaire responses measuring demographics, shopping motivations, and individual differences. Using this data, an exploratory machine learning approach was utilised to investigate the demographic and psychological predictors of frequent plastic bag consumption. It was found that frequent bag buyers spent more money in store, were younger, more likely to be male, less frugal, open to new experiences, and more displeased with their appearance (compared with infrequent bag buyers). Interestingly, environmental concerns did not predict plastic bag consumption, highlighting the disconnect between predicting pro-environmental attitudes and real world environmental behaviour.

Forty million people are estimated to be in some form of modern slavery across the globe. Understanding the factors that make any particular individual or geographical region vulnerable to such abuse is essential for the development of effective interventions and policy. Efforts to isolate and assess the importance of individual drivers statistically are impeded by two key challenges: data scarcity and high dimensionality, typical of many “wicked problems”. The hidden nature of modern slavery restricts available data points; and the large number of candidate variables that are potentially predictive of slavery inflate the feature space exponentially. The result is a “small n, large p” setting, where overfitting and significant inter-correlation of explanatory variables can render more traditional statistical approaches problematic. Recent advances in non-parametric computational methods, however, offer scope to overcome such challenges and better capture the complex nature of modern slavery. We present an approach that combines non-linear machine-learning models and strict cross-validation methods with novel variable importance techniques, emphasising the importance of stability of model explanations via a Rashomon-set analysis. This approach is used to model the prevalence of slavery in 48 countries, with results bringing to light the importance of new predictive factors—such as a country’s capacity to protect the physical security of women, which has been previously under-emphasised in quantitative models. Further analyses uncover that women are particularly vulnerable to exploitation in areas where there is poor access to resources. Our model was then leveraged to produce new out-of-sample estimates of slavery prevalence for countries where no survey data currently exists.