BIO3097 Assessed Coursework

Monte Carlo Pertemuan 9

MDB RE finance

Entrepreneurship education programmes are often evaluated using pre-post designs that focus on changes in learner outcomes over time. However, such evaluations may overlook the role of baseline positioning in shaping observed impacts. This study examines whether learners participating in the Wavumbuzi Entrepreneurship Challenge (WEC) are entering with progressively higher baseline entrepreneurial competencies, and whether this may explain the declining number of statistically significant improvements observed in recent cohorts, particularly in Kenya. Using matched pre-post data from six WEC editions across Kenya and Rwanda, the analysis compares baseline scores and outcome trajectories across 18 entrepreneurial factors, including entrepreneurial experience, intentions, mindset constructs, and applied competencies. Baseline trends are analysed longitudinally, and changes are assessed using adjusted difference scores that account for reverse-coded variables, alongside statistical significance and effect size measures. The findings reveal a clear upward trend in baseline scores for applied entrepreneurial competencies, such as opportunity recognition, problem-solving, and taking action, across both countries. In Kenya, these increases are more concentrated and coincide with a reduction in the number of competencies showing statistically significant improvement in the most recent edition (WEC-KE-6). This pattern suggests the presence of a ceiling effect, where higher starting points limit the scope for measurable gains. In contrast, Rwanda exhibits more balanced baseline growth across both applied and selected motivational competencies, with more consistent improvements across recent cohorts.

Election monitoring increasingly depends on the capacity to process high-volume citizen reports, social media posts, and platform-based submissions under conditions of uncertainty. This study evaluates whether a zero-shot large language model (LLM) pipeline can function as a high-recall triage system for election monitoring using VoteReportPH data from the 2025 Philippine elections. Drawing on signal detection theory, information overload theory, and human-AI complementarity, the study frames LLM classification as decision support rather than autonomous adjudication. The analysis used a postprocessed Election Monitoring System dataset of 3,618 reports and a cleaned model-evaluation dataset of 4,158 reports. For binary validity detection, the model correctly surfaced 166 of 181 human-validated reports, yielding a recall of 0.9171, specificity of 0.8973, and accuracy of 0.8983, but low precision of 0.3198. This error profile indicates a recall-oriented filter that reduces missed incidents while forwarding false positives for human review. In multiclass incident categorization, the model performed strongly on explicit categories such as automated counting machine errors and illegal campaigning, but weakly on rare, residual, and procedurally ambiguous categories. The findings show that zero-shot LLMs can support civic monitoring as triage infrastructure, but they require human verification, transparent error handling, and category-specific workflow design.