Computational and specialist methods for research questions that go beyond standard statistics: prediction and classification, large text collections, relational and spatial data, and quantitative synthesis of published studies.
These methods share a need for careful validation. Machine learning models must be evaluated on data they were not trained on, with attention to overfitting, class imbalance and interpretability. Text mining and sentiment analysis depend on preprocessing choices and on validating automated labels against human judgment. Network analysis and GIS work bring their own data structures — ties between actors, coordinates and spatial dependence — which standard tests ignore.
Meta-analysis sits here because it is a specialist statistical method in its own right: computing comparable effect sizes, choosing fixed or random-effects models, quantifying heterogeneity and checking for publication bias. If you are still planning the review that will feed a meta-analysis, the research services pillar covers systematic review design and searching.
Support for research-grade machine learning: framing the problem, preventing leakage, choosing and tuning models with proper cross-validation, interpreting them with tools like SHAP, and reporting transparently.
Computational text analysis for research: corpus preparation, topic modeling with LDA, STM or BERTopic, sentiment and classification with lexicons or transformer models, and validation you can report.
Help turning relational data into a network, choosing measures that answer your question, detecting communities, visualizing clearly and, where needed, testing network hypotheses statistically.
Support for research GIS: preparing and projecting spatial data, making accurate maps, testing spatial autocorrelation, finding hot spots and fitting spatial regression or geographically weighted models.
Statistical support for meta-analysis: extracting and converting effect sizes, fitting fixed or random-effects models, exploring heterogeneity, assessing small-study effects and reporting to PRISMA 2020.
Tell us what you’re working on and where you’re stuck. We’ll recommend the right kind of help and send a quote.