Corrected Akaike Information Criterion Research Articles

Public health insurance, frailty, and lack of home support predict rehab discharge following elective anterior cervical discectomy and fusion

BACKGROUND CONTEXTAnterior cervical discectomy and fusion (ACDF) is a commonly-performed and generally well-tolerated procedure used to treat cervical disc herniation. Rarely, patients require discharge to inpatient rehab, leading to inconvenience for the patient and increased healthcare expenditure for the medical system. PURPOSEThe objective of this study was to create an accurate and practical predictive model for, as well as delineate associated factors with, rehab discharge following elective ACDF. STUDY DESIGNThis was a retrospective, single-center, cohort study. PATIENT SAMPLEPatients who underwent ACDF between 2012 and 2022 were included. Those with confounding diagnoses or who underwent concurrent, staged, or nonelective procedures were excluded. OUTCOME MEASURESPrimary outcomes for this study included measurements of accuracy for predicting rehab discharge. Secondary outcomes included associations of variables with rehab discharge. METHODSCurrent Procedural Terminology codes identified patients. Charts were reviewed to obtain additional demographic and clinical characteristics on which an initial univariate analysis was performed. Two logistic regression and two machine learning models were trained and evaluated on the data using cross-validation. A multimodel logistic regression was implemented to analyze independent variable associations with rehab discharge. RESULTSA total of 466 patients were included in the study. The logistic regression model with minimum corrected Akaike information criterion score performed best overall, with the highest values for area under the receiver operating characteristic curve (0.83), Youden's J statistic (0.71), balanced accuracy (85.7%), sensitivity (90.3%), and positive predictive value (38.5%). Rehab discharge was associated with a modified frailty index of 2 (p=.007), lack of home support (p=.002), and having Medicare or Medicaid insurance (p=.007) after correction for multiple hypotheses. CONCLUSIONSNonmedical social determinants of health, such as having public insurance or a lack of support at home, may play a role in rehab discharge following elective ACDF. In combination with the modified frailty index and other variables, these factors can be used to predict rehab discharge with high accuracy, improving the patient experience and reducing healthcare costs.

Extended Kumaraswamy Exponential Distribution with Application to COVID-19 Data set

There are many probability models describing the time related events data. In this study, the exponential distribution is modified by adding one more parameter to get more flexible probability model called Extended Kumaraswamy Exponential (EKwE) distribution using the New Kw-G family (NKwG) of distributions. We have studied some of the statistical characteristics of the model, such as its reliability function, hazard rate function, and quantile function. For testing the applicability of the model, a real data set based on COVID-19 data is taken. The Cramer-von Mises (CVM) approach, Least Square Estimation (LSE), and Maximum Likelihood Estimation (MLE) are used to estimate the model’s parameters. Validity of the model is checked by using P-P plot and Q-Q plot. Akaike Information Criterion (AIC), Corrected Akaike Information Criterion (CAIC), Bayesian Information Criterion (BIC) and Hannan-Quinn Information Criterion (HQIC) are also used for model comparison. Goodness of fit of the proposed model is tested using Kolmogrov-Smirnov (KS), Cramer-Von Mises (CVM) and Anderson-Darling (An) test statistics along with respective p-values. All the analysis of the study is performed by using R programming.

Modeling and Forecasting Inflation in Nigeria: A Time Series Regression with ARIMA Method

This study uses time series regression with autoregressive integrated moving average (ARIMA) modeling to establish a model for forecasting inflation in Nigeria for the period 1981-2020. Akaike Information Criterion Corrected (AICC) and Bayesian Information Criterion (BIC) were used to select the best model among competing models. Through these methods, regression with ARIMA (0,0,1) error was selected as the most parsimonious model for inflation forecasting in Nigeria. The results of the out-sample-forecast show that a high inflation rate will be experienced by the end of 2023, and between 2024 and 2030, the inflation rate will be alternating but will maintain a lower rate than that of 2023.

Comparison between the Four-parameter and Five-parameter Logistics Equations in the Determination of the Limits of Detection of Ochratoxin A

Ochratoxin A, a type of mycotoxin, was detected in plant sample matrices using the "optical waveguide lightmode spectroscopy" (OWLS) method. The calibration curve for the detection of ochratoxin A utilizing "optical waveguide lightmode spectroscopy" (OWLS) displayed a sigmoidal shape; hence, the 5-parameter logistics (5-PL) or 4-parameters logistics (4-PL) model should be used to fit the data rather than a linear model. Through the use of error function analysis and various functions such as Bias Factor, corrected Akaike Information Criterion (AICc), adjusted correlation coefficient (adjR2), Hannan–Quinn information criterion (HQC), Root Mean Square Error (RMSE), Bayesian Information Criterion (BIC), and Accuracy Factor, the distinction between the 5-PL and 4-PL models is found to be inconsistent. The half maximal effective concentration (EC50) confidence intervals overlapped, indicating no difference between the two methods; the 4-PL model was chosen because of its fewer parameters. The Limits of Detection (LOD) calculated using the 4-PL equation was 0.818 ng/mL, with a 95% confidence interval of 0.526-1.143 ng/mL. This value falls within the estimated range of 0.5 to 10 ng/mL found in the first investigation. Not only was the linear component of the data curve successfully represented by this study's use of the 4-PL model, but the model was also successful in representing the entire date curve.

Nodal Electricity Price Forecasting using Exponential Smoothing and Holt’s Exponential Smoothing

The prediction of nodal electricity price (NEP) is a primary step to be done before the bidding process starts in the actual market environment. NEP plays a significant role for the efficient working of the electrical system. NEP follows a common trend as during peak hours when the load is high the price will also be high similarly during off-peak-load times the price will be lower and common to all the node. Thus, accurate forecasting of the NEP can help electricity generation companies to be more proactive in the wholesale electricity market to maximize its overall benefits. In this paper, exponential smoothing (ES), and holt’s exponential smoothing (HES) have been utilized for forecasting the NEP. Furthermore, a comparative analysis between ES and HES has been done considering several alpha values and several trends. The model evaluation and the forecasting performance have been tested using different parameters of ES, and HES techniques such as Akaike Information Criterion (AIC), Akaike Information Criterion Corrected (AICc), Bayesian Information Criteria (BIC). The performance of the proposed technique has been authenticated efficaciously on average nodal real-time price data collected from ISO New England (BOSTON Zone).

Prevalence of HIV in Kazakhstan 2010-2020 and Its Forecasting for the Next 10 Years.

HIV is a growing public health burden that threatens thousands of people in Kazakhstan. Countries around the world, including Kazakhstan, are facing significant problems in predicting HIV infection prevalence. It is crucial to understand the epidemiological trends of infectious diseases and to monitor the prevalence of HIV in a long-term perspective. Thus, in this study, we aimed to forecast the prevalence of HIV in Kazakhstan for 10 years from 2020 to 2030 by using mathematical modeling and time series analysis. We use statistical Autoregressive Integrated Moving Average (ARIMA) models and a nonlinear epidemic Susceptible-Infected (SI) model to forecast the HIV infection prevalence rate in Kazakhstan. We estimated the parameters of the models using open data on the prevalence of HIV infection among women and men (aged 15-49 years) in Kazakhstan provided by the Kazakhstan Bureau of National Statistics. We also predict the effect of pre-exposure prophylaxis (PrEP) control measures on the prevalence rate. The ARIMA (1,2,0) model suggests that the prevalence of HIV infection in Kazakhstan will increase from 0.29 in 2021 to 0.47 by 2030. On the other hand, the SI model suggests that this parameter will increase to 0.60 by 2030 based on the same data. Both models were statistically significant by Akaike Information Criterion corrected (AICc) score and by the goodness of fit. HIV prevention under the PrEP strategy on the SI model showed a significant effect on the reduction of the HIV prevalence rate. This study revealed that ARIMA (1,2,0) predicts a linear increasing trend, while SI forecasts a nonlinear increase with a higher prevalence of HIV. Therefore, it is recommended for healthcare providers and policymakers use this model to calculate the cost required for the regional allocation of healthcare resources. Moreover, this model can be used for planning effective healthcare treatments.

Analisis Faktor-Faktor yang Menjelaskan Kasus AIDS Provinsi Jawa Timur Menggunakan Model Geographically Weighted Logistic Regression (GWLR)

AIDS is the most chronic phase of HIV infection which can weaken the immune system. In 2020, East Java Province is a province which has the most HIV infections and in the third place for the highest total number of AIDS cases in Indonesia. The purpose of this research is to build a model using Geographically Weighted Logistic Regression (GWLR), and to work out the grouping results of regencies/cities using K-means Clustering Analysis. The variables used in this research are Gini Ratio, L Index of Per Capita Expenditure, Gender Ratio, Dependency Ratio, Gender Development Index, and The Number of Pos Pelayanan KB Desa. The proportion levels of AIDS cases are categorized into 2 categories based on cut-point which has been specified, which 0 as the category of low level with the proportion of AIDS cases is less than 0.0006 and 1 as the category of high level with the proportion of AIDS cases is more than or equal to 0.0006. Parameter estimation for GWLR is using Maximum Likelihood Estimation (MLE) method with Fixed Gaussian as weighted kernel function and optimum bandwidth is determined using Akaike's Information Criterion Corrected (AICc). Z-Score of the most suitable model will be grouped using K-means Clustering Analysis, with Z-score is parameter estimator divided by standard error. Grouping results indicates cluster 1 members tend to be regencies/cities that have gender ratio and dependency ratio as significant variables, meanwhile cluster 2 members tend to be regencies/cities that have only dependency ratio as significant variable.

Comparison between Modified Weighted Pareto Distribution and Many other Distributions

In 2020 one of the researchers in this paper, in his first research, tried to find out the Modified Weighted Pareto Distribution of Type I by using the Azzalini method for weighted distributions, which contain three parameters, two of them for scale while the third for shape.This research compared the distribution with two other distributions from the same family; the Standard Pareto Distribution of Type I and the Generalized Pareto Distribution by using the Maximum likelihood estimator which was derived by the researchers for Modified Weighted Pareto Distribution of Type I, then the Mont Carlo method was used–that is one of the simulation manners for generating random samples data in different sizes ( n= 10,30,50), and in different initial values for each Pareto distribution family being used in the research. A comparison was done by using Akaike Information Criteria, Corrected Akaike Information Criteria, and Bayesian Information Criteria

Top-down and bottom-up effects and relationships with local environmental factors in the water frog–helminth systems in Latvia

Semi-aquatic European water frogs (Pelophylax spp.) harbour rich helminth infra-communities, whose effects on host population size in nature are poorly known. To study top-down and bottom-up effects, we conducted calling male water frog counts and parasitological investigations of helminths in waterbodies from different regions of Latvia, supplemented by descriptions of waterbody features and surrounding land use data. We performed a series of generalized linear model and zero-inflated negative binomial regressions to determine the best predictors for frog relative population size and helminth infra-communities. The highest-ranked (by Akaike information criterion correction, AICc) model explaining the water frog population size contained only waterbody variables, followed by the model containing only land use within 500 m, while the model containing helminth predictors had the lowest rank. Regarding helminth infection responses, the relative importance of the water frog population size varied from being non-significant (abundances of larval plagiorchiids and nematodes) to having a similar weight to waterbody features (abundances of larval diplostomids). In abundances of adult plagiorchiids and nematodes the best predictor was the host specimen size. Environmental factors had both direct effects from the habitat features (e.g., waterbody characteristics on frogs and diplostomids) and indirect effects through parasite-host interactions (impacts of anthropogenic habitats on frogs and helminths). Our study suggests the presence of synergy between top-down and bottom-up effects in the water frog–helminth system that creates a mutual dependence of frog and helminth population sizes and helps to balance helminth infections at a level that does not cause over-exploitation of the host resource.

Applying geographically weighted regression to quantify the impact of impervious surface density on land surface temperature in Ho Chi Minh City, Vietnam

Surface characteristics are believed to be the primary factor causing spatial variation in land surface temperature (LST), especially in a large and rapidly growing city like Ho Chi Minh City. The study aims to estimate the spatial relationship between land surface temperature (LST) and impervious surface density (ISA) by using three statistical techniques: 1) Optimized hot spot analysis; 2) Global linear regression (OLS); and 3) Geographically weighted regression (GWR). LST and ISA were extracted from Landsat 8 OLI of 2020. According to the optimized hot spot analysis’s findings, over 80% of the areas of high ISA values (mean ISA: 0.86, SD: 0.13) overlap with areas of high LST values (mean LST: 36.2°C, SD: 2.04°C), with a correlation coefficient of r=0.775. Likewise, Moran’s I index (I=0.82, Z=72, P<0.001) indicates that LST in Ho Chi Minh City has a significantly spatial dependence, so geographically weighted regression (GWR) and global linear (OLS) models need to be taken into account simultaneously. Coefficient of determination R2 and AICc (Akaike information criterion corrected) were used to compare the models. R2 increases from 0.78 to 0.87, whereas AICc decreases from 8047 to 7033 in OLS and GWR, respectively. GWR has a more substantial explanatory power for the association between LST and ISA and offers superior LST prediction accuracy compared to OLS. Potential urban heat island areas can be accurately detected by combining GWR and optimized hot spot analysis.

Analysis and Modeling Gross Domestic Product, Carbon Dioxide Emission, Population Growth, and Life Expectancy at Birth: Case Study in Qatar

Studies on Life Expectancy at Birth (LEB) and Population Growth (PG) with several economic variables for cases in Qatar have not been carried out much. The study of the relationship between Life expectancy at Birth(LEB), Carbon Dioxide(CO2)emission, Population growth(PG), and Gross Domestic Product(GDP)for the case of Qatar is interesting because it is a developed country based on oil and gas. The aims of this study is to build a dynamic model for vector time series that describes the relationship between the variables discussed. From the analysis results based on Criterion Information, Akaike's Information Criterion Correction (AICC), and Cointegration Rank, the best model obtained is the Vector Error Correction Model with order 4 (VECM (4)) and with Cointegration Rank r=4. Based on this dynamic model, Granger-Causality analysis, Impulse Response Function (IRF), and Forecasting will be discussed.

Synthesis and Performance Evaluation of Novel Bentonite-Supported Nanoscale Zero Valent Iron for Remediation of Arsenic Contaminated Water and Soil

Groundwater arsenic (As) pollution is a naturally occurring phenomenon posing serious threats to human health. To mitigate this issue, we synthesized a novel bentonite-based engineered nano zero-valent iron (nZVI-Bento) material to remove As from contaminated soil and water. Sorption isotherm and kinetics models were employed to understand the mechanisms governing As removal. Experimental and model predicted values of adsorption capacity (qe or qt) were compared to evaluate the adequacy of the models, substantiated by error function analysis, and the best-fit model was selected based on corrected Akaike Information Criterion (AICc). The non-linear regression fitting of both adsorption isotherm and kinetic models revealed lower values of error and lower AICc values than the linear regression models. The pseudo-second-order (non-linear) fit was the best fit among kinetic models with the lowest AICc values, at 57.5 (nZVI-Bare) and 71.9 (nZVI-Bento), while the Freundlich equation was the best fit among the isotherm models, showing the lowest AICc values, at 105.5 (nZVI-Bare) and 105.1 (nZVI-Bento). The adsorption maxima (qmax) predicted by the non-linear Langmuir adsorption isotherm were 354.3 and 198.5 mg g-1 for nZVI-Bare and nZVI-Bento, respectively. The nZVI-Bento successfully reduced As in water (initial As concentration = 5 mg L-1; adsorbent dose = 0.5 g L-1) to below permissible limits for drinking water (10 µg L-1). The nZVI-Bento @ 1% (w/w) could stabilize As in soils by increasing the amorphous Fe bound fraction and significantly diminish the non-specific and specifically bound fraction of As in soil. Considering the enhanced stability of the novel nZVI-Bento (upto 60 days) as compared to the unmodified product, it is envisaged that the synthesized product could be effectively used for removing As from water to make it safe for human consumption.

Screening main and interaction effects in a Plackett-Burman design

In a supersaturated design, the number of factors is greater than the number of observations. A design that is not supersaturated but has more parameters to be estimated than observations results in a supersaturated model (SSM). An SSM may occur when two-factor interactions exist in addition to main effects. The goal is to find the few active effects from the many in the SSM. Several approaches may be used to find a set of potential models and then to select the best model using some criterion. This may be done in a single stage by considering all main effects and two-factor interactions simultaneously, or by screening main effects first, then identifying the active interactions. In this paper, we study a single-stage and multiple two-stage approaches as well as different algorithms and criteria for selecting the models. For the Plackett-Burman design we study, we find that the SCAD algorithm with the corrected Akaike Information criterion (cAIC) generally performs well, but the best procedure depends on the true, unknown, underlying model.

Enhanced bifunctional adsorption of anionic and cationic pollutants by MgAl LDH nanosheets modified montmorillonite via acid-salt activation

A novel H+ and Cl− intercalated montmorillonite@MgAl LDH (MMT-N) nanocomposite was prepared by a one-step hydrothermal method following acid-salt treatment. The MMT-N prepared by the advanced modification method contains low CO2–3contamination, high specific surface area, and high porosity. By enhancing the ion exchange of montmorillonite and LDH, efficient removal of anionic and anionic dyes was achieved without the introduction of another cation contamination. The adsorption kinetics of methyl orange (MO) and methylene blue (MB) are in accordance with the pseudo-second-order model, and the maximum adsorption amounts of methylene blue and methyl orange reached 815.998 mg g−1 and 399.836 mg g−1, respectively. The coefficient of determination (R2) and the corrected Akaike Information Criterion (AICc) revealed that the adsorption isotherm data of MO and MB on MMT-80 fitted well with Freundlich isotherm, signifying that the adsorption process was dominated by chemisorption. And the adsorption of MO and MB by MMA-80 was endothermic and exothermic, respectively. The “intercalation-split mechanism” and “ion exchange” mechanisms of MO and MB in MMA-80 were also investigated. This study provides an efficient method for the synthesis of high-performance bifunctional adsorbents for cationic and anionic pollutants and offers potential applications for practical wastewater purification.

A novel long-short term memory network approach for stress model updating for excavations in high stress environments

ABSTRACT Digitalisation has increased access to large amounts of data for rock engineers. Machine learning presents an opportunity to aid data interpretation. The operators of Garson Mine use a microseismic database to calibrate a mine-scale finite difference model, used to assess seismic risk to inform mine operations. A Long-Short Term Memory (LSTM) network is proposed for stress model updating. The model is trained using microseismic data, geology, and geomechanical parameters from the FLAC3D model. Two LSTM networks are developed for Garson Mine: (1) predicting far field principal stresses in the FLAC3D model, and (2) predicting the far field six-component stress tensors in the model. Various LSTM network hyperparameters were analyzed to determine the architecture for the targets: input encoding and pre-processing, training solver, network layer architecture, and cost function. Architectures were chosen based on the corrected Akaike Information Criterion (AICc), coefficient of determination (R2), and percent capture (%C). When predicting principal stresses, AICc = −59.62, R2 = 0.996, and %C = 97%, and when predicting the six-component stress tensor AICc = −45.50, R2 = 0.997, and %C = 80%. This research represents progress towards continuous, automated updating of numerical models such that rapid, more accurate forecasts of changes in stress conditions will allow earlier reaction to challenging stress environments, increasing safety of excavations.

Modeling Rainfall Intensity-Duration-Frequency (IDF) and Establishing Climate Change Existence in Uyo-Nigeria Using Non-Stationary Approach

This study aims at establishing if climate change exists in the Niger Delta environment using non-stationary rainfall Intensity-Duration-Frequency (IDF) modelling incorporating time-variant parameters. To compute the intensity levels, the open-access R-studio software was used based on the General Extreme Value (GEV) distribution function. Among the four linear parameter models adopted for integrating time as a covariate, the fourth linear model incorporating scale and location with the shape function constant produced the least corrected Akaike Information Criteria (AICc), varying between 306.191 to 101.497 for 15 and 1440 minutes, respectively, selected for calibration of the GEV distribution equation. The non-stationary intensities yielded higher values above those of stationary models, proving that the assumption of stationary IDF models underestimated extreme events. The difference of 13.71 mm/hr (22.71%) to 14.26 mm/hr (17.0%) intensities implies an underestimation of the peak flood from a stationary IDF curve. The statistical difference at a 95% confidence level between stationary and non-stationary models was significant, confirming evidence of climatic change influenced by time-variant parameters. Consequently, emphasis should be on applying shorter-duration storms for design purposes occurring with higher intensities to help reduce the flood risk and resultant infrastructural failures.

Modelling the Kinetics of Tartrazine Sorption by Bottom Ash

Bottom ash is the solid residue left over from municipal waste combustion or incineration in a Municipal Waste Incineration Furnace. Its use as a sorption agent, particularly for dye sorption, is a new and important application. Linearized adsorption kinetics has drawbacks such as inaccurate representation of the parameters' 95 percent confidence interval output, unbalanced attention to potential outliers, and magnification of errors may result in inaccurate parameter values. In this study, we used nonlinear regression to investigate 16 adsorption kinetics models of tartrazine by bottom ash. The pseudo-second order was the best model based on the Bias and Accuracy factor near unity, but based on other error function analysis, this model performs equally well with the exponential and fractal-like pseudo-second order based other error functions such as Root-Mean-Square Error (RMSE), adjusted coefficient of determination (adjR2), Marquardt’s percent standard deviation (MPSD), Bayesian Information Criterion (BIC), Hannan-Quinn Information Criterion (HQC), and especially the corrected Akaike Information Criterion (AICc) function as the absolute difference is 5 absolute unit making discriminatory activity difficult. Furthermore, because the pseudo-second order and exponential models have only two parameters, they are less complicated according to Occam's razor. Because the pseudo-second order model is more popular and has more applications than the less well-known exponential model, we chose it as the best model for tartrazine sorption to bottom ash. Kinetic analysis using the PSO model gave a value of equilibrium adsorption capacity, qe of 21.88 mg g-1 (95% confidence interval (C.I.), 20.93 to 22.84) and k2 (g/(mg.sec)) of 0.00002 (95%, C.I., 0.00001 to 0.00002).

Modelling the Kinetics of Tartrazine Sorption by the Rhizopus arrhizus Biomass

Rhizopus arrhizus is utilized in many biotechnological applications such as in the manufacturing of enzymes, including pectinases, amylases, proteases, cellulases, and phytases, and metabolites such as lactic acid, ethanol and fumaric acids. Its spent biomass is useful in the food and feed industry. Its usage as a biosorption agent especially dye sorption is an emerging and important application. In this study we explore 16 adsorption kinetics model of tartrazine by R. arrhizus using nonlinear regression. Based on the statistical indicators especially penalty-based error functions such as adjusted coefficient of determination (R2), Root-Mean-Square Error (RMSE), corrected Akaike Information Criterion (AICc), Bayesian Information Criterion (BIC), Hannan-Quinn Information Criterion (HQC) and Marquardt's percentage standard deviation (MPSD) shows that the pseudo-2nd order (PSO) was the best model followed by pseudo-nth order and Fractal-like Pseudo-2nd Order. The parameter of the PSO model gave a value of equilibrium adsorption capacity, qe of 9.367 mg g-1 (95% confidence interval (C.I.), 9.250 to 9.485) and k2 (g/(mg.min)) of 0.037 (95%, C.I., 0.032 to 0.041). The nonlinear regression exercise allows the uncertainty determination of the parameters to be carried out.

Comparison of two non-linear functions describing the growth of Popielno White and New Zealand White rabbits

The objective of this study was to compare growth curves of Popielno White and New Zealand White rabbits modelled by the von Bertalanffy and Gompertz functions. There were 37 Popielno White (21 males and 16 females) and 55 New Zealand White (28 males and 27 females) rabbits used. The nonlinear regression procedure (NLIN) in the SAS 9.4 software (SAS, 2014) was used for modelling rabbits growth curves from birth to 12th week of age. To see how the process of rabbit growth might run in next weeks, we extrapolated weights of rabbits up to 30th week of age. Six criteria: the mean error, mean squared error, mean absolute error, quotient between the error sum of squares and the observed sum of squares, corrected Akaike Information Criterion and Bayesian Information Criterion were used to compare models. The shape of all curves was sigmoid and both models well fitted to data from birth to 12th week of age, i.e. during interpolation process. The problem with the fit occurred later, during extrapolation process, i.e. when models were used to predict the course of the growth of animals in the next weeks of age. When body weight was extrapolated in later weeks of rabbits life, we noticed slightly worse fitting of von Bertalanffy model and overestimation of the asymptotic body weight of animals. To predict the further course of rabbit growth in practice, the Gompertz function would be a better choice. Additionally the Gompertz model could be a helpful tool for breeders to describe growth process of rabbits and to select a more economically profitable breed.

Uncovering drivers of community-level house price dynamics through multiscale geographically weighted regression: A case study of Wuhan, China

For buyers, investors and urban policy, understanding drivers of community-level house prices across space and across time, are important for urban management and economic planning. In this study, we interrogated two housing market datasets, one from 2015, the other from 2019, for Wuhan, China, in order to uncover diversities and similarities in the spatial relationships between house price and contextual data; and in the context of increasingly volatile markets. A non-stationary approach was adopted with basic geographically weighted regression (GWR) and multiscale GWR (MGWR), where only the latter enables relationships to vary at their own spatial scale. In terms of model fit, both MGWR (adj. R2: 0.94 and 0.97, for 2015 and 2019, respectively) and GWR (adj. R2: 0.87 and 0.81) represented an improvement over the usual linear regression (adj. R2: 0.11 and 0.09) and the spatial lag mode (adj. R2: 0.21 and 0.27). Similarly marked improvements for GWR and for MGWR were found using corrected Akaike Information Criterion (AICc) based fit diagnostics. However, of more importance and via MGWR, the spatially varying drivers of house price were found to operate at a range of spatial scales, that in turn changed in strength and significance between the two study years. Such insights allow for spatially- and temporally-aware decision- and policy-making for housing price control and urban planning, given China’s housing markets can be increasing prone to strong growth coupled with severe depressions.