Drought remains one of the most significant climate-related hazards in the Philippines, particularly in Central Luzon, where agriculture, water resources, and local livelihoods are highly sensitive to rainfall variability. This study characterized the spatiotempo-ral behavior of meteorological drought in Central Luzon and evaluated forecasting ap-proaches for short- and long-range drought prediction using the Standardized Precipi-tation Index (SPI).
Monthly Climate Hazards Group InfraRed Precipitation with Station Data (CHIRPS) rainfall data from 1981–2025 were quality-controlled and validated against Philippine Atmospheric, Geophysical and Astronomical Services Administration (PAGASA) sta-tion observations. Following validation, SPI was computed to assess drought variability, identify spatial drought hotspots, evaluate long-term trends, and examine spatial clus-tering patterns. Two forecasting frameworks were then developed for SPI-3 prediction: a Conditional Long Short-Term Memory (LSTM) model and a Bayesian spatiotempo-ral model incorporating Besag–York–Mollie´ 2 (BYM2) spatial random effects, AR(1) temporal dependence, climatic forcing, and seasonal predictors.
Results revealed pronounced spatial heterogeneity in drought occurrence across Central Luzon, with persistent drought hotspots identified in portions of Nueva Ecija, Bulacan, Tarlac, and Zambales. These areas consistently exhibited higher drought fre-quency and greater vulnerability than the rest of the region.
Overall, the findings indicate that both the Conditional LSTM and Bayesian spa-tiotemporal models are suitable for short-term drought forecasting, with the Conditional LSTM providing slightly higher predictive accuracy at shorter lead times. For medium-and long-range forecasting, the Bayesian framework exhibited more stable predictive
v

vi

performance while offering the additional advantages of model interpretability and un-certainty quantification, making it well suited for drought outlooks and decision-support applications.

This study constructs a daily Philippine financial system sentiment index using a hierarchical dynamic factor model (HDFM) estimated via the Kalman expectation-maximization algorithm. The index synthesizes 22 indicators spanning the banking, equities, fixed income, and external segments of the Philippine financial system into a composite system-wide sentiment measure and four sectoral sub-indices, covering the period January 2016 to March 2026. The modeling framework accommodates the mixed-frequency and ragged-edge structure of the indicator panel — with observation frequencies ranging from daily to quarterly. The extracted composite index captures the two most significant stress episodes within the sample — the COVID-19 onset in March 2020 and the 2022 global monetary tightening cycle. The sentiment index represents the first application of the Moench–Ng–Potter hierarchical factor framework to an emerging market financial system and constitutes a novel daily-frequency macroprudential surveillance tool for the Bangko Sentral ng Pilipinas.

The field of systematic review and meta-analysis allows researchers to combine multiple studies to obtain a more precise estimate of clinical phenomena. This research focuses on determining reference intervals, which provides a range of val-ues used by clinicians to classify subjects as healthy or not. Existing methods for estimating reference intervals in meta-analysis rely on strict parametric assump-tions, which lead to severe variance overestimation and artificially inflated interval lengths when study populations are skewed or heterogeneous. To resolve these limitations, this study proposes a nonparametric methodology based on Kernel Density Estimation integrated into a Mixture Distribution framework (KDE-MD) under a fixed-effects model, the preferred approach when only a few studies are available. By superimposing individual density curves rather than pooling their variances, the KDE-MD method successfully adapts to complex, mixed data struc-tures without requiring pre-specified distribution families. Furthermore, because non-parametric density estimation inherently risks finite-sample over-smoothing, a bootstrap coverage calibration procedure is introduced to dynamically adjust the nominal confidence level. Extensive simulation results demonstrate that the Bootstrap-Calibrated KDE-MD (BootKDE-MD) methodology effectively detects and corrects these finite-sample biases. Ultimately, the proposed method consis-tently guarantees the accurate nominal target coverage while yielding narrower interval lengths than traditional parametric alternatives.

In laboratory medicine, reference ranges are essential tools to determine whether an individual’s laboratory measurements of biochemical quantities are typical of that of a healthy individual. While reference ranges are commonly used to assess one analyte, more complex diagnosis of diseases requires the examination of two or more analytes. Traditional methods to account for these correlations yield ellipsoidal reference regions, whose shape makes them difficult to interpret. To improve the interpretability of these reference regions, several researchers have developed rectangular reference regions. This current study aims to improve the construction of rectangular reference regions by offering novel methodologies for their computation through the use of Bayesian prediction regions under both the covariate-free and covariate-dependent settings. The use of a Bayesian approach through improper priors allows direct sampling from the posterior predictive distribution, addressing the limitation of the frequentist approach that rely on numerical estimation in computing the unknown predictive factor. Simulations were performed to evaluate the proposed reference regions in terms of their accuracy and precision. Results show that the proposed regions perform comparable to the Bonferroni simultaneous regions in terms of accuracy, and are more precise for a minimum sample size of 100 when the desired region is two- or three-dimensional.

Multidimensional poverty considers deprivation in different aspects of well-being. Despite existing global and national multidimensional poverty index (MPI) methodologies, gaps remain in its measurement, particularly in incorporating spatial information and generating finer disaggregation. To come up with spatially varying weights, the study employed two PCA methods adapted for spatial analysis, namely, Geographically Weighted Principal Component Analysis (GWPCA) and PCA with Spatial Autocorrelation (SPCA). Furthermore, two small area estimation techniques, namely, Regression synthetic estimation and Empirical Best Linear Unbiased Prediction (EBLUP), were applied to generate small area MPI estimates.
Incidence of deprivation was observed to be the highest for educational attainment at
84.90 percent. This reflects the close relationship between education and poverty, where poor educational outcomes often lead to limited economic opportunities. Results of MPI using nested equal weights, GWPCA-, and SPCA-based weights do not differ much at the national level. The headcount ratio, or the proportion of the population who are multidimensionally poor, was estimated at around 20 to 21 percent. Meanwhile, the intensity of poverty, or the average deprivation score of multidimensionally poor people, was estimated at 39 to 45 percent. Lastly, MPI, which captures both headcount ratio and intensity of poverty, was estimated at 8 to 9.5 percent. These results were found to be consistent with official poverty statistics, providing evidence to the interconnectedness of multidimensional poverty and income poverty.
Across the three weighting schemes, the province of Sulu consistently had the highest estimates of headcount ratio, intensity, and MPI. Meanwhile, the province of Pampanga consistently had the lowest estimates for the same measures. For the estimation of city/municipality level MPI, Regression synthetic estimation using weights based on SPCA yielded the highest number of valid estimates among all approaches. Consistent with the results of the provincial level MPI, majority of the ten municipalities with the highest MPI were from the province of Sulu. The combination of SPCA-based weights and Regression synthetic estimation emerged as the most effective approach for estimating multidimensional poverty at finer geographic levels.
Keywords: multidimensional poverty index, geographically weighted PCA, PCA with spatial autocorrelation, regression synthetic estimation, EBLUP