Lyson Chaka

Dr. Lyson Chaka

Statistics, Department of Science and Technology, Pretoria, Gauteng, South Africa
University of Pretoria, South Africa

Research Focus & Metrics

Academic Career Length

5Years (2022-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
15 July 2025 in Frontiers in Applied Mathematics and Statistics
Verified
Journal Article
Open access
PDF file
AAAlbert Antwi
Albert Antwi
EKEmelia Thembile Kammies
Emelia Thembile Kammies
Lyson ChakaLyson Chaka
Lyson Chaka
MAMartins Akugbe Arasomwan
Martins Akugbe Arasomwan
Introduction: Accurate price forecasts and the evaluation of some of the factors that affect the prices of grains are crucial for proper planning and food security. Various methods have been designed to model and forecast grain prices and other time-stamped data. However, due to some inherent limitations, some of these models do not produce accurate forecasts or are not easily interpretable. Although dynamic Bayesian generalized additive models (GAMs) offer potential to overcome some of these problems, they do not explicitly model local trends. This may lead to biased fixed effects estimates and forecasts, thus highlighting a significant gap in literature.Methods: To address this, we propose the use of random intercepts to capture localized trends within the dynamic Bayesian GAM framework to forecast South African wheat and maize prices. Furthermore, we examine the complex underlying relationships of the prices with inflation and rainfall.Results: Evidence from the study suggests that the proposed method is able to adequately capture the dynamic localized trends consistent with the underlying local trends in the prices. It was observed that the estimated localized variations are significant, which led to improved and efficient fixed-effect parameter estimates. This led to better posterior predictions and forecasts. A comparison to the static trend Bayesian GAMs and the autoregressive integrated moving average (ARMA) models indicates a general superiority of the proposed approach for the posterior predictions and long-term posterior forecasts and has potential for short-term forecasts. The static trend Bayesian GAMs were found to generally outperform the ARMA models in long-term posterior forecasts and also have potential for short-term forecasts. However, for 1-step ahead posterior forecasts, the ARMA models consistently outperformed all the Bayesian models. The study also unveiled a significant direct nonlinear impact of inflation on wheat and maize prices. Although the impacts of rainfall on wheat and maize prices are indirect and nonlinear, only the impact on maize prices is significant.Discussion: The improved efficiency and forecasts of our proposed method suggest that researchers and practitioners may consider the approach when modelling and forecasting long-term prices of grains, other agricultural commodities, speculative assets and general single-subject time series data exhibiting non-stationarity.

6 May 2022 in Stats
Verified
Journal Article
Open access
PDF file
Lyson ChakaLyson Chaka
Lyson Chaka
PNPeter Njuho
Peter Njuho
The design and analysis of experiments which involve factors each consisting of both fixed and random levels fit into linear mixed models. The assumed linear mixed-model design matrix takes either a full-rank or less-than-full-rank form. The complexity of the data structures of such experiments falls in the model-selection and parameter-estimation process. The fundamental consideration in the estimation process of linear models is the special case in which elements of the error vector are assumed equal and uncorrelated. However, different assumptions on the structure of the variance–covariance matrix of error vector in the estimation of parameters of a linear mixed model may be considered. We conceptualise a repeated-measures design with multiple between-subjects factors, in which each of these factors has both fixed and random levels. We focus on the construction of linear mixed-effects models, the estimation of variance components, and hypothesis testing in which the default covariance structure of homoscedastic error terms is not appropriate. We illustrate the proposed approach using longitudinal data fitted to a three-factor linear mixed-effects model. The novelty of this approach lies in the exploration of the fixed and random levels of the same factor and in the subsequent interaction effects of the fixed levels. In addition, we assess the differences between levels of the same factor and determine the proportion of the total variation accounted for by the random levels of the same factor.

27 August 2025 in Scientific African
Verified
Journal Article
Open access
Lyson ChakaLyson Chaka
Lyson Chaka
Mohamed A. M. Abd ElbasitMohamed A. M. Abd Elbasit
Mohamed A. M. Abd Elbasit
SJSimbarashe Jombo
Simbarashe Jombo


1 April 2026 in The Journal of Agricultural Science
Verified
Journal Article
Closed access
PDF file
Lyson ChakaLyson Chaka
Lyson Chaka
PNPeter Njuho
Peter Njuho
HPHans-Peter Piepho
Hans-Peter Piepho
JMJean Claude Malela-Majika
Jean Claude Malela-Majika
Experimental designs involving factors with a mix of fixed and random levels have been explored by few. These designs are useful when comparing a set of new treatments (fixed levels) to a population of established treatments (random levels). This approach enables partitioning variability and testing of both fixed and random effects, leading to improved estimates and more reliable inference. However, combining analyses of variance from partitioned data poses challenges, including data rearrangement, methodology, and coding complexities. Existing statistical software does not directly support combined analyses for such designs. The current study provides guidelines for conducting combined analysis of variance in linear mixed-effects models where factors have both fixed and random levels. The approach utilizes SAS PROC GLIMMIX and tools from the Comprehensive R Archive Network to compile the combined analysis of variance, followed by multiple comparisons of treatment means. The procedure is demonstrated using treatment structures in a completely randomized design, split-split-plot design, and repeated-measures design. The method can be extended to other experimental designs. This framework addresses a critical gap in existing literature by providing a practical and readily applicable analytical tool that enables integrated analysis across all factor levels within a single linear mixed-model structure, while accounting for pre-history effects associated with prior management practices in complex experimental systems.

25 May 2026 in Communications in Statistics - Theory and Methods
Verified
Journal Article
Open access
BMBusisiwe Masilela
Busisiwe Masilela
JMJean Claude Malela-Majika
Jean Claude Malela-Majika
Lyson ChakaLyson Chaka
Lyson Chaka