This work-in-progress encompasses a structured and systematic approach to handling complex and nested parameter structures in R, using the 'plot.matrix' function as an example. Internally, plot.matrix decomposes into a series of base R plotting functions ('plot', 'polygon', 'text', 'axis') and exposes a range of configurable arguments. While powerful, this interface can become verbose, rigid and difficult to manage, particularly when working with nested or partially specified parameters.

To address this issue, we prototype a more modular interface based on the 'argList' package that captures function arguments as programmable values. This allows us to capture both unevaluated and evaluated function arguments in a programmable form.

Our approach is centered on three complementary strategies: parameter propagation, which expands shorthand inputs into structured lists; global parameter injection, which distributes shared parameters to nested components; and local parameter injection, which supports fine-grained control using a dot-notation syntax (e.g., 'plot1.type' → 'plot1$type').

These mechanisms are implemented via utility functions ('propagate_param', 'global_param', and 'local_param') that facilitate clear and priority-aware argument resolution. The current implementation is demonstrated through basic examples, which are intentionally minimal to highlight the internal logic rather than practical use cases.

Although still in early stages, this framework is designed for generalization and reuse, particularly in scenarios where function interfaces require flexibility, nesting, or extensibility. The long-term goal is to enable more concise, maintainable, and user-friendly APIs in R, applicable not only in graphical workflows but in any domain involving complex parameter management. Future work will focus on practical applications and interface validation.

Understanding monetary policy has always been of paramount economic and political importance. However, it remains a difficult task, despite transparency efforts and the regular flow of information to the public, which becomes even more complex when communication channels are multilingual. This paper examines the policy narratives of the Swiss National Bank (SNB) in terms of language and policy preferences, using the corpus of speeches delivered by its members over the period 1997-2022. Using a dynamic semantic search strategy based on top2vec, the framework analysis was able to identify interlingual similarities and differences with the help of pre-trained multilingual models. The results show that the SNB's communication strategy is strongly oriented towards the objectives assigned to the central bank, with attention being paid to systemic risks, banking regulation and financial markets, which emerge as second but no less important objectives, closely linked to the international environment, in particular the Eurosystem as a strategic aspect of the stability of the Swiss franc. The results suggest that English is used exclusively to address core central banking issues (monetary policy, inflation and interest rates), while uncertainty concerns seem to be reported more in German or French. The resulting dual semantic space, consisting of embedded words and documents, yielded relevant topics with respect to the size and scope of the corpus. Furthermore, informative indices could be constructed for policy measurement, as a crisis index was found to be consistent with the business cycle fluctuations and technical recessions experienced in Switzerland over the last 25 years.

Understanding how voters transition between parties is central to post-election analysis, particularly amid recent rightward shifts in electoral outcomes. Beyond traditional polling-based methods, voter transition matrices can be estimated from aggregate election results using ecological inference methods. Previous approaches primarily employed hierarchical Bayesian models, while recent work has shifted to constrained optimization techniques. Both approaches perform well but face challenges, notably the underlying assumption that voter transition behaviour is homogeneous across electoral districts. We regard this assumption as critical and propose a more flexible framework for modelling voter transitions.

To relax the homogeneity assumption, we introduce a mixture-model approach, whereby electoral districts are grouped into clusters, that is mixture components, exhibiting similar voter movement patterns. We treat the number of mixture components as a hyperparameter to be determined separately.

Estimation is carried out using a stochastic Expectation-Maximization algorithm, which proves to be numerically flexible. As result we obtain the cluster specific transition matrices and for each voting district a (posterior) distribution for the mixture components. The stochastic nature of the EM algorithm allows to assess the estimation variability of the mixture components. In particular, we can assess the variability of the membership probabilities, which can be used to quantify our confidence in each assignment.

We illustrate our method with an analysis of the 299 electoral districts from the 2025 German Federal Election and present the resulting voter movement estimates and clusters.

Estimating the Complier Average Causal Effect (CACE) in instrumental variable (IV) settings is critical for uncovering causal relationships, especially when treatment compliance varies across subpopulations. In practice, policy interventions often suffer from imperfect compliance and regions of covariate space where the overlap assumption is violated—challenges that undermine the reliability of standard IV estimation methods. This work extends the Bayesian Additive Regression Trees with Instrumental Variables (BART-IV) framework to a (Transformed) Random Forest-IV setting to improve both flexibility and computational efficiency in CACE estimation.

We introduce a novel methodology that incorporates kernel-based weighting to balance observable covariates between instrument-defined groups, thereby addressing overlap violations and mitigating problems from near-deterministic instrument assignment probabilities. Our contributions are threefold:

1. We reinterpret the BART-IV framework as a general two-step procedure, enabling the use of alternative machine learning models, such as Random Forests, in place of BART for CACE estimation.

2. We develop a Random Forest-IV approach that offers competitive or superior performance relative to BART-IV, particularly when the binary covariate assumption is relaxed.

3. We integrate kernel-based weights into the transformed Random Forest-IV and GRF-IV framework, improving robustness in settings with extreme propensity scores.

Through simulation studies, we demonstrate that our approach maintains high estimation accuracy across varying degrees of treatment effect heterogeneity. The kernel-weighted extension is especially effective in stabilizing estimates when propensity scores approach 0 or 1, conditions that often lead to extreme inverse probability weights.